| 1 | /* |
| 2 | |
| 3 | Derby - Class org.apache.derby.impl.store.raw.log.LogToFile |
| 4 | |
| 5 | Copyright 1997, 2004 The Apache Software Foundation or its licensors, as applicable. |
| 6 | |
| 7 | Licensed under the Apache License, Version 2.0 (the "License"); |
| 8 | you may not use this file except in compliance with the License. |
| 9 | You may obtain a copy of the License at |
| 10 | |
| 11 | http://www.apache.org/licenses/LICENSE-2.0 |
| 12 | |
| 13 | Unless required by applicable law or agreed to in writing, software |
| 14 | distributed under the License is distributed on an "AS IS" BASIS, |
| 15 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 16 | See the License for the specific language governing permissions and |
| 17 | limitations under the License. |
| 18 | |
| 19 | */ |
| 20 | |
| 21 | package org.apache.derby.impl.store.raw.log; |
| 22 | |
| 23 | import org.apache.derby.iapi.services.diag.Performance; |
| 24 | |
| 25 | import org.apache.derby.impl.store.raw.log.CheckpointOperation; |
| 26 | import org.apache.derby.impl.store.raw.log.LogCounter; |
| 27 | import org.apache.derby.impl.store.raw.log.LogRecord; |
| 28 | import org.apache.derby.impl.store.raw.log.StreamLogScan; |
| 29 | |
| 30 | // need this to print nested exception that corrupts the database |
| 31 | import org.apache.derby.iapi.services.context.ErrorStringBuilder; |
| 32 | |
| 33 | import org.apache.derby.iapi.services.info.ProductGenusNames; |
| 34 | import org.apache.derby.iapi.services.info.ProductVersionHolder; |
| 35 | |
| 36 | import org.apache.derby.iapi.reference.MessageId; |
| 37 | import org.apache.derby.iapi.reference.Property; |
| 38 | import org.apache.derby.iapi.reference.SQLState; |
| 39 | |
| 40 | import org.apache.derby.iapi.services.daemon.DaemonService; |
| 41 | import org.apache.derby.iapi.services.daemon.Serviceable; |
| 42 | import org.apache.derby.iapi.services.context.ContextManager; |
| 43 | import org.apache.derby.iapi.services.context.ContextService; |
| 44 | import org.apache.derby.iapi.services.context.ShutdownException; |
| 45 | import org.apache.derby.iapi.services.monitor.Monitor; |
| 46 | import org.apache.derby.iapi.services.monitor.ModuleControl; |
| 47 | import org.apache.derby.iapi.services.monitor.ModuleSupportable; |
| 48 | import org.apache.derby.iapi.services.monitor.PersistentService; |
| 49 | import org.apache.derby.iapi.services.sanity.SanityManager; |
| 50 | import org.apache.derby.iapi.services.io.Formatable; |
| 51 | import org.apache.derby.iapi.services.io.TypedFormat; |
| 52 | import org.apache.derby.iapi.services.io.FormatIdUtil; |
| 53 | import org.apache.derby.iapi.services.io.StoredFormatIds; |
| 54 | import org.apache.derby.iapi.services.stream.HeaderPrintWriter; |
| 55 | import org.apache.derby.iapi.services.stream.PrintWriterGetHeader; |
| 56 | import org.apache.derby.iapi.services.stream.InfoStreams; |
| 57 | import org.apache.derby.iapi.error.StandardException; |
| 58 | import org.apache.derby.iapi.services.i18n.MessageService; |
| 59 | import org.apache.derby.iapi.store.access.AccessFactory; |
| 60 | import org.apache.derby.iapi.store.access.AccessFactoryGlobals; |
| 61 | import org.apache.derby.iapi.store.access.TransactionController; |
| 62 | import org.apache.derby.iapi.store.raw.Loggable; |
| 63 | import org.apache.derby.iapi.store.raw.RawStoreFactory; |
| 64 | import org.apache.derby.iapi.store.raw.ScanHandle; |
| 65 | import org.apache.derby.iapi.store.raw.log.LogFactory; |
| 66 | import org.apache.derby.iapi.store.raw.log.Logger; |
| 67 | import org.apache.derby.iapi.store.raw.log.LogInstant; |
| 68 | import org.apache.derby.iapi.store.raw.log.LogScan; |
| 69 | import org.apache.derby.iapi.store.raw.Transaction; |
| 70 | import org.apache.derby.iapi.store.raw.xact.RawTransaction; |
| 71 | import org.apache.derby.iapi.store.raw.xact.TransactionFactory; |
| 72 | import org.apache.derby.iapi.store.raw.data.DataFactory; |
| 73 | import org.apache.derby.iapi.services.property.PersistentSet; |
| 74 | |
| 75 | import org.apache.derby.iapi.store.access.DatabaseInstant; |
| 76 | import org.apache.derby.catalog.UUID; |
| 77 | import org.apache.derby.iapi.services.uuid.UUIDFactory; |
| 78 | import org.apache.derby.iapi.services.property.PropertyUtil; |
| 79 | import org.apache.derby.iapi.reference.Attribute; |
| 80 | import org.apache.derby.iapi.services.io.FileUtil; |
| 81 | import org.apache.derby.iapi.util.ReuseFactory; |
| 82 | |
| 83 | import org.apache.derby.io.StorageFactory; |
| 84 | import org.apache.derby.io.WritableStorageFactory; |
| 85 | import org.apache.derby.io.StorageFile; |
| 86 | import org.apache.derby.io.StorageRandomAccessFile; |
| 87 | |
| 88 | import java.io.File; // Plain files are used for backups |
| 89 | import java.io.IOException; |
| 90 | import java.io.SyncFailedException; |
| 91 | import java.io.ByteArrayOutputStream; |
| 92 | import java.io.DataOutputStream; |
| 93 | import java.io.ByteArrayInputStream; |
| 94 | import java.io.DataInputStream; |
| 95 | import java.io.FileNotFoundException; |
| 96 | |
| 97 | import java.net.MalformedURLException; |
| 98 | import java.net.URL; |
| 99 | |
| 100 | import java.util.Properties; |
| 101 | import java.util.Vector; |
| 102 | import java.util.zip.CRC32; |
| 103 | |
| 104 | /** |
| 105 | |
| 106 | This is an implementation of the log using a non-circular file system file. |
| 107 | No support for incremental log backup or media recovery. |
| 108 | Only crash recovery is supported. |
| 109 | <P> |
| 110 | The 'log' is a stream of log records. The 'log' is implemented as |
| 111 | a series of numbered log files. These numbered log files are logically |
| 112 | continuous so a transaction can have log records that span multiple log files. |
| 113 | A single log record cannot span more then one log file. The log file number |
| 114 | is monotonically increasing. |
| 115 | <P> |
| 116 | The log belongs to a log factory of a RawStore. In the current implementation, |
| 117 | each RawStore only has one log factory, so each RawStore only has one log |
| 118 | (which composed of multiple log files). |
| 119 | At any given time, a log factory only writes new log records to one log file, |
| 120 | this log file is called the 'current log file'. |
| 121 | <P> |
| 122 | A log file is named log<em>logNumber</em>.dat |
| 123 | <P> |
| 124 | Everytime a checkpoint is taken, a new log file is created and all subsequent |
| 125 | log records will go to the new log file. After a checkpoint is taken, old |
| 126 | and useless log files will be deleted. |
| 127 | <P> |
| 128 | RawStore exposes a checkpoint method which clients can call, or a checkpoint is |
| 129 | taken automatically by the RawStore when |
| 130 | <OL> |
| 131 | <LI> the log file grows beyond a certain size (configurable, default 100K bytes) |
| 132 | <LI> RawStore is shutdown and a checkpoint hasn't been done "for a while" |
| 133 | <LI> RawStore is recovered and a checkpoint hasn't been done "for a while" |
| 134 | </OL> |
| 135 | <P> |
| 136 | This LogFactory is responsible for the formats of 2 kinds of file: the log |
| 137 | file and the log control file. And it is responsible for the format of the |
| 138 | log record wrapper. |
| 139 | <P> <PRE> |
| 140 | |
| 141 | Format of log control file |
| 142 | |
| 143 | @format_id FILE_STREAM_LOG_FILE |
| 144 | @purpose The log control file contains information about which log files |
| 145 | are present and where the last checkpoint log record is located. |
| 146 | @upgrade |
| 147 | @disk_layout |
| 148 | (pre-v15) |
| 149 | int format id |
| 150 | int log file version |
| 151 | long the log instant (LogCounter) of the last completed checkpoint |
| 152 | (v15 onward) |
| 153 | int format id |
| 154 | int obsolete log file version |
| 155 | long the log instant (LogCounter) of the last completed checkpoint |
| 156 | int JBMS version |
| 157 | int checkpoint interval |
| 158 | long spare (value set to 0) |
| 159 | long spare (value set to 0) |
| 160 | long spare (value set to 0) |
| 161 | |
| 162 | @end_format |
| 163 | </PRE> |
| 164 | <HR WIDTH="100%"> |
| 165 | <PRE> |
| 166 | |
| 167 | Format of the log file |
| 168 | |
| 169 | @format_id FILE_STREAM_LOG_FILE |
| 170 | @purpose The log file contains log record which record all the changes |
| 171 | to the database. The complete transaction log is composed of a series of |
| 172 | log files. |
| 173 | @upgrade |
| 174 | @disk_layout |
| 175 | int format id - the format Id of this log file |
| 176 | int obsolete log file version - not used |
| 177 | long log file number - this number orders the log files in a |
| 178 | series to form the complete transaction log |
| 179 | long prevLogRecord - log instant of the previous log record, in the |
| 180 | previous log file. |
| 181 | |
| 182 | [log record wrapper]* one or more log records with wrapper |
| 183 | |
| 184 | int endMarker - value of zero. The beginning of a log record wrapper |
| 185 | is the length of the log record, therefore it is never zero |
| 186 | [int fuzzy end]* zero or more int's of value 0, in case this log file |
| 187 | has been recovered and any incomplete log record set to zero. |
| 188 | @end_format |
| 189 | </PRE> |
| 190 | <HR WIDTH="100%"> |
| 191 | <PRE> |
| 192 | |
| 193 | Format of the log record wrapper |
| 194 | |
| 195 | @format_id none. The format is implied by the FILE_STREAM_LOG_FILE |
| 196 | @purpose The log record wrapper provides information for the log scan. |
| 197 | @upgrade |
| 198 | @disk_layout |
| 199 | length(int) length of the log record (for forward scan) |
| 200 | instant(long) LogInstant of the log record |
| 201 | logRecord(byte[length]) byte array that is written by the FileLogger |
| 202 | length(int) length of the log record (for backward scan) |
| 203 | @end_format |
| 204 | </PRE> |
| 205 | |
| 206 | |
| 207 | <P>Multithreading considerations:<BR> |
| 208 | Log Factory must be MT-safe. |
| 209 | <P> |
| 210 | Class is final as it has methods with privilege blocks |
| 211 | and implements PrivilegedExceptionAction. |
| 212 | */ |
| 213 | |
| 214 | public final class LogToFile implements LogFactory, ModuleControl, ModuleSupportable, |
| 215 | Serviceable, java.security.PrivilegedExceptionAction |
| 216 | { |
| 217 | |
| 218 | private static int fid = StoredFormatIds.FILE_STREAM_LOG_FILE; |
| 219 | |
| 220 | // format Id must fit in 4 bytes |
| 221 | |
| 222 | /** |
| 223 | Return my format identifier. |
| 224 | */ |
| 225 | public int getTypeFormatId() { |
| 226 | return StoredFormatIds.FILE_STREAM_LOG_FILE; |
| 227 | } |
| 228 | |
| 229 | // at the beginning of every log file is the following information: |
| 230 | // the log file formatId |
| 231 | // the log file version (int) |
| 232 | // the log file number (long) |
| 233 | // the log instant at the end of the last log record in the previous file (long) |
| 234 | public static final int LOG_FILE_HEADER_SIZE = 24; |
| 235 | |
| 236 | protected static final int LOG_FILE_HEADER_PREVIOUS_LOG_INSTANT_OFFSET = LOG_FILE_HEADER_SIZE-8; |
| 237 | |
| 238 | // Number of bytes overhead of each log record. |
| 239 | // 4 bytes of length at the beginning, 8 bytes of log instant, |
| 240 | // and 4 bytes of length at the end |
| 241 | public static final int LOG_RECORD_OVERHEAD = 4+8+4; |
| 242 | |
| 243 | public static final String DBG_FLAG = SanityManager.DEBUG ? "LogTrace" : null; |
| 244 | public static final String DUMP_LOG_ONLY = SanityManager.DEBUG ? "DumpLogOnly" : null; |
| 245 | public static final String DUMP_LOG_FROM_LOG_FILE = |
| 246 | SanityManager.DEBUG ? "derby.storage.logDumpStart" : null; |
| 247 | protected static final String LOG_SYNC_STATISTICS = "LogSyncStatistics"; |
| 248 | |
| 249 | // If you change this number, then JBMS 1.1x and 1.2x will give a really |
| 250 | // horrendous error message when booting against a db created by you. When |
| 251 | // we decided that we don't need to worry about people mis-using the |
| 252 | // product that way, then we can change this. Just remember, before we do, |
| 253 | // all existing database will have the number 9 in there. |
| 254 | private static final int OBSOLETE_LOG_VERSION_NUMBER = 9; |
| 255 | |
| 256 | /* how big the log file should be before checkpoint or log switch is taken */ |
| 257 | private static final int DEFAULT_LOG_SWITCH_INTERVAL = 1024*1024; |
| 258 | private static final int LOG_SWITCH_INTERVAL_MIN = 100000; |
| 259 | private static final int LOG_SWITCH_INTERVAL_MAX = 128*1024*1024; |
| 260 | private static final int CHECKPOINT_INTERVAL_MIN = 100000; |
| 261 | private static final int CHECKPOINT_INTERVAL_MAX = 128*1024*1024; |
| 262 | private static final int DEFAULT_CHECKPOINT_INTERVAL = 10*1024*1024; |
| 263 | |
| 264 | //log buffer size values |
| 265 | private static final int DEFAULT_LOG_BUFFER_SIZE = 32768; //32K |
| 266 | private static final int LOG_BUFFER_SIZE_MIN = 8192; //8k |
| 267 | private static final int LOG_BUFFER_SIZE_MAX = LOG_SWITCH_INTERVAL_MAX; |
| 268 | private int logBufferSize = DEFAULT_LOG_BUFFER_SIZE; |
| 269 | |
| 270 | /* Log Control file flags. */ |
| 271 | private static final byte IS_BETA_FLAG = 0x1; |
| 272 | |
| 273 | /** |
| 274 | * When the derby.system.durability property is set to 'test', the store |
| 275 | * system will not force sync calls in the following cases |
| 276 | * - for the log file at each commit |
| 277 | * - for the log file before data page is forced to disk |
| 278 | * - for page allocation when file is grown |
| 279 | * - for data writes during checkpoint |
| 280 | * This means it is possible that the recovery system may not work properly, |
| 281 | * committed transactions may be lost, and/or database may not |
| 282 | * be in a consistent state. |
| 283 | * In order that we recognize this case that the database was previously |
| 284 | * at any time booted in this mode, this value is written out |
| 285 | * into the log control file. This will help prevent us from |
| 286 | * wasting time to resolve issues in such cases. |
| 287 | * @see org.apache.derby.iapi.reference.Property#DURABILITY_PROPERTY |
| 288 | * This value is written as part of the log control file flags byte. |
| 289 | */ |
| 290 | private static final byte IS_DURABILITY_TESTMODE_NO_SYNC_FLAG = 0x2; |
| 291 | |
| 292 | /** |
| 293 | * keeps track of if the database was booted previously at any time with |
| 294 | * derby.system.durability=test |
| 295 | */ |
| 296 | private static boolean wasDBInDurabilityTestModeNoSync = false; |
| 297 | |
| 298 | /* to err on the conservative side, unless otherwise set, assume log |
| 299 | * archive is ON |
| 300 | */ |
| 301 | private static final String DEFAULT_LOG_ARCHIVE_DIRECTORY = "DEFAULT"; |
| 302 | |
| 303 | private int logSwitchInterval = DEFAULT_LOG_SWITCH_INTERVAL; |
| 304 | private int checkpointInterval = DEFAULT_CHECKPOINT_INTERVAL; |
| 305 | |
| 306 | String dataDirectory; // where files are stored |
| 307 | private WritableStorageFactory logStorageFactory; |
| 308 | |
| 309 | private boolean logBeingFlushed; // is the log in the middle of a flush |
| 310 | // (access of the variable should sync on this) |
| 311 | |
| 312 | protected LogAccessFile logOut; // an output stream to the log file |
| 313 | // (access of the variable should sync on this) |
| 314 | private StorageRandomAccessFile firstLog = null; |
| 315 | protected long endPosition = -1; // end position of the current log file |
| 316 | long lastFlush = 0; // the position in the current log |
| 317 | // file that has been flushed to disk |
| 318 | |
| 319 | long logFileNumber = -1; // current log file number |
| 320 | // other than during boot and recovery time, |
| 321 | // logFileNumber is only changed by |
| 322 | // switchLogFile, which is synchronized. |
| 323 | // |
| 324 | // MT - Anyone accessing this number should |
| 325 | // synchronized on this if the current log file |
| 326 | // must not be changed. If not synchronized, |
| 327 | // the log file may have been switched. |
| 328 | |
| 329 | long firstLogFileNumber = -1; |
| 330 | // first log file that makes up the active |
| 331 | // portion (with active transactions) of the |
| 332 | // log. |
| 333 | // |
| 334 | // MT - This value is set during recovery or |
| 335 | // during log truncation. In the former single |
| 336 | // thread is assumed. In the latter |
| 337 | // must be synchronized with this to access |
| 338 | // or change. |
| 339 | |
| 340 | private long maxLogFileNumber = LogCounter.MAX_LOGFILE_NUMBER; |
| 341 | private CheckpointOperation currentCheckpoint; |
| 342 | // last checkpoint successfully taken |
| 343 | // |
| 344 | // MT - only changed or access in recovery or |
| 345 | // checkpoint, both are single thread access |
| 346 | |
| 347 | long checkpointInstant; |
| 348 | // log instant of te curerntCheckpoint |
| 349 | |
| 350 | private DaemonService checkpointDaemon; // the background worker thread who is going to |
| 351 | // do checkpoints for this log factory. |
| 352 | |
| 353 | private int myClientNumber; |
| 354 | // use this number to talk to checkpoint Daemon |
| 355 | |
| 356 | private volatile boolean checkpointDaemonCalled; |
| 357 | // checkpoint Daemon called already - it is not |
| 358 | // important that this value is correct, the |
| 359 | // daemon just need to be called once in a |
| 360 | // while. Deamon can handle multiple posts. |
| 361 | |
| 362 | private long logWrittenFromLastCheckPoint = 0; |
| 363 | // keeps track of the amout of log written between checkpoints |
| 364 | private RawStoreFactory rawStoreFactory; |
| 365 | // use this only after recovery is finished |
| 366 | |
| 367 | protected DataFactory dataFactory; |
| 368 | // use this only after revocery is finished |
| 369 | |
| 370 | protected boolean ReadOnlyDB; // true if this db is read only, i.e, cannot |
| 371 | // append log records |
| 372 | |
| 373 | |
| 374 | // DEBUG DEBUG - do not truncate log files |
| 375 | private boolean keepAllLogs; |
| 376 | |
| 377 | // if database is encrypted, the content of the log files are encrypted |
| 378 | private boolean databaseEncrypted; |
| 379 | |
| 380 | // the following booleans are used to put the log factory into various |
| 381 | // states |
| 382 | private boolean recoveryNeeded = true; // log needs to be recovered |
| 383 | private boolean inCheckpoint = false; // in the middle of a checkpoint |
| 384 | private boolean inRedo = false; // in the middle of redo loop |
| 385 | private boolean inLogSwitch = false; |
| 386 | |
| 387 | // make sure we don't do anything after the log factory has been stopped |
| 388 | private boolean stopped = false; |
| 389 | |
| 390 | // if log is to go to another device, this variable is set. If null, then |
| 391 | // log goes to the log subdirectory underneath the data directory |
| 392 | String logDevice; |
| 393 | |
| 394 | // disable syncing of log file when running in derby.system.durability=test |
| 395 | private boolean logNotSynced = false; |
| 396 | |
| 397 | private volatile boolean logArchived = false; |
| 398 | private boolean logSwitchRequired = false; |
| 399 | |
| 400 | /** DEBUG test only */ |
| 401 | int test_logWritten = 0; |
| 402 | int test_numRecordToFillLog = -1; |
| 403 | private int mon_flushCalls; |
| 404 | private int mon_syncCalls; |
| 405 | private int mon_numLogFlushWaits; |
| 406 | private boolean mon_LogSyncStatistics; |
| 407 | private int mon_numBytesToLog; |
| 408 | |
| 409 | |
| 410 | /** |
| 411 | If not null then something is corrupt in the raw store and this represents the original error. |
| 412 | */ |
| 413 | protected volatile StandardException corrupt; |
| 414 | |
| 415 | /** |
| 416 | If frozen, don't allow anything on disk to change. |
| 417 | */ |
| 418 | private boolean isFrozen; |
| 419 | |
| 420 | /** |
| 421 | Product Version information. Invarient after boot. |
| 422 | */ |
| 423 | ProductVersionHolder jbmsVersion; |
| 424 | |
| 425 | /** |
| 426 | On disk database version information. When running in soft upgrade this version |
| 427 | may be different to jbmsVersion. |
| 428 | */ |
| 429 | private int onDiskMajorVersion; |
| 430 | private int onDiskMinorVersion; |
| 431 | private boolean onDiskBeta; |
| 432 | |
| 433 | private CRC32 checksum = new CRC32(); // holder for the checksum |
| 434 | |
| 435 | |
| 436 | /** |
| 437 | * Note: Why logging system support file sync and write sync ? |
| 438 | * Note : The reason to support file and write sync of logs is |
| 439 | * there was no support to do write sync until jdk1.4 and then |
| 440 | * there was write sync jvm bug in jdk1.4.1, only in jdk1.4.2 write |
| 441 | * sync(rws mode) mechanism can be used corretly. |
| 442 | * Default in JVMS >= jdk1.4.2 is write sync(see the boot method for jvm checks). |
| 443 | * |
| 444 | * Write sync mechanism support is added for performance reasons. |
| 445 | * On commits, logging system has to make sure the log for committed |
| 446 | * transaction is on disk. With out write sync , log is written to the |
| 447 | * disk and then fsync() is used on commits to make log is written to the |
| 448 | * disk for sure. On most of the OS , fsync() calls are expensive. |
| 449 | * On heavey commit oriented systems , file sync make the system run slow. |
| 450 | * This problem is solved by using write sync on preallocated log file. |
| 451 | * write sync is much faster than doing write and file sync to a file. |
| 452 | * File should be preallocated for write syncs to perform better than |
| 453 | * the file sync method. Whenever a new log file is created, |
| 454 | * logSwitchInterval size is preallocated by writing zeros after file after the header. |
| 455 | */ |
| 456 | |
| 457 | /*If set to true , write sync will be used to do log write other file |
| 458 | * level sync is used. |
| 459 | */ |
| 460 | private boolean isWriteSynced = false; |
| 461 | |
| 462 | |
| 463 | // log file that is yet to be copied to backup, updates to this variable |
| 464 | // needs to visible checkpoint thread. |
| 465 | private volatile long logFileToBackup ; |
| 466 | // It is set to true when online backup is in progress, updates to |
| 467 | // this variable needs to visible to checkpoint thread. |
| 468 | private volatile boolean backupInProgress = false; |
| 469 | |
| 470 | |
| 471 | /** |
| 472 | MT- not needed for constructor |
| 473 | */ |
| 474 | public LogToFile() { |
| 475 | keepAllLogs = PropertyUtil.getSystemBoolean(RawStoreFactory.KEEP_TRANSACTION_LOG); |
| 476 | |
| 477 | |
| 478 | if (Performance.MEASURE) |
| 479 | mon_LogSyncStatistics = PropertyUtil.getSystemBoolean(LOG_SYNC_STATISTICS); |
| 480 | } |
| 481 | |
| 482 | /* |
| 483 | ** Methods of Corruptable |
| 484 | */ |
| 485 | |
| 486 | /** |
| 487 | Once the log factory is makred as corrupt then the raw sto |
| 488 | */ |
| 489 | public StandardException markCorrupt(StandardException originalError) { |
| 490 | |
| 491 | boolean firsttime = false; |
| 492 | |
| 493 | synchronized (this) |
| 494 | { |
| 495 | if (corrupt == null && originalError != null) |
| 496 | { |
| 497 | corrupt = originalError; |
| 498 | firsttime = true; |
| 499 | } |
| 500 | } |
| 501 | |
| 502 | // only print the first error |
| 503 | if (corrupt == originalError) |
| 504 | logErrMsg(corrupt); |
| 505 | |
| 506 | |
| 507 | // this is the first time someone detects error, shutdown the |
| 508 | // system as much as possible without further damaging it |
| 509 | if (firsttime) |
| 510 | { |
| 511 | synchronized(this) |
| 512 | { |
| 513 | stopped = true; |
| 514 | |
| 515 | if (logOut != null) |
| 516 | { |
| 517 | try |
| 518 | { |
| 519 | logOut.corrupt(); // get rid of open file descriptor |
| 520 | } |
| 521 | catch (IOException ioe) |
| 522 | { |
| 523 | // don't worry about it, just trying to clean up |
| 524 | } |
| 525 | } |
| 526 | |
| 527 | // NullPointerException is preferred over corrupting the database |
| 528 | logOut = null; |
| 529 | } |
| 530 | |
| 531 | if (dataFactory != null) |
| 532 | dataFactory.markCorrupt(null); |
| 533 | |
| 534 | } |
| 535 | |
| 536 | return originalError; |
| 537 | } |
| 538 | |
| 539 | private void checkCorrupt() throws StandardException |
| 540 | { |
| 541 | synchronized (this) |
| 542 | { |
| 543 | if (corrupt != null) |
| 544 | { |
| 545 | throw StandardException.newException( |
| 546 | SQLState.LOG_STORE_CORRUPT, corrupt); |
| 547 | } |
| 548 | } |
| 549 | } |
| 550 | |
| 551 | /* |
| 552 | ** Methods of LogFactory |
| 553 | */ |
| 554 | |
| 555 | /** |
| 556 | MT- not needed |
| 557 | */ |
| 558 | public Logger getLogger() { |
| 559 | |
| 560 | if (ReadOnlyDB) |
| 561 | return null; |
| 562 | else |
| 563 | return new FileLogger(this); |
| 564 | } |
| 565 | |
| 566 | /** |
| 567 | Recover the rawStore to a consistent state using the log. |
| 568 | |
| 569 | <P> |
| 570 | In this implementation, the log is a stream of log records stored in |
| 571 | one or more flat files. Recovery is done in 2 passes: redo and undo. |
| 572 | <BR> <B>Redo pass</B> |
| 573 | <BR> In the redo pass, reconstruct the state of the rawstore by |
| 574 | repeating exactly what happened before as recorded in the log. |
| 575 | <BR><B>Undo pass</B> |
| 576 | <BR> In the undo pass, all incomplete transactions are rolled back in |
| 577 | the order from the most recently started to the oldest. |
| 578 | |
| 579 | <P>MT - synchronization provided by caller - RawStore boot. |
| 580 | This method is guaranteed to be the only method being called and can |
| 581 | assume single thread access on all fields. |
| 582 | |
| 583 | @see Loggable#needsRedo |
| 584 | @see FileLogger#redo |
| 585 | |
| 586 | @exception StandardException Standard Cloudscape error policy |
| 587 | */ |
| 588 | public void recover( |
| 589 | RawStoreFactory rsf, |
| 590 | DataFactory df, |
| 591 | TransactionFactory tf) |
| 592 | throws StandardException |
| 593 | { |
| 594 | if (SanityManager.DEBUG) |
| 595 | { |
| 596 | SanityManager.ASSERT(rsf != null, "raw store factory == null"); |
| 597 | SanityManager.ASSERT(df != null, "data factory == null"); |
| 598 | } |
| 599 | |
| 600 | checkCorrupt(); |
| 601 | |
| 602 | rawStoreFactory = rsf; |
| 603 | dataFactory = df; |
| 604 | |
| 605 | // initialize the log writer only after the rawstorefactory is available, |
| 606 | // log writer requires encryption block size info from rawstore factory |
| 607 | // to encrypt checksum log records. |
| 608 | if (firstLog != null) |
| 609 | logOut = new LogAccessFile(this, firstLog, logBufferSize); |
| 610 | |
| 611 | // we don't want to set ReadOnlyDB before recovery has a chance to look |
| 612 | // at the latest checkpoint and determine that the database is shutdown |
| 613 | // cleanly. If the medium is read only but there are logs that need |
| 614 | // to be redone or in flight transactions, we are hosed. The logs that |
| 615 | // are redone will leave dirty pages in the cache. |
| 616 | |
| 617 | if (recoveryNeeded) |
| 618 | { |
| 619 | try |
| 620 | { |
| 621 | ///////////////////////////////////////////////////////////// |
| 622 | // |
| 623 | // During boot time, the log control file is accessed and |
| 624 | // logFileNumber is determined. LogOut is not set up. |
| 625 | // LogFileNumber is the log file the latest checkpoint lives in, |
| 626 | // or 1. It may not be the latest log file (the system may have |
| 627 | // crashed between the time a new log was generated and the |
| 628 | // checkpoint log written), that can only be determined at the |
| 629 | // end of recovery redo. |
| 630 | // |
| 631 | ///////////////////////////////////////////////////////////// |
| 632 | |
| 633 | FileLogger logger = (FileLogger)getLogger(); |
| 634 | |
| 635 | ///////////////////////////////////////////////////////////// |
| 636 | // |
| 637 | // try to find the most recent checkpoint |
| 638 | // |
| 639 | ///////////////////////////////////////////////////////////// |
| 640 | if (checkpointInstant != LogCounter.INVALID_LOG_INSTANT) |
| 641 | { |
| 642 | currentCheckpoint = |
| 643 | findCheckpoint(checkpointInstant, logger); |
| 644 | } |
| 645 | |
| 646 | // if we are only interested in dumping the log, start from the |
| 647 | // beginning of the first log file |
| 648 | if (SanityManager.DEBUG) |
| 649 | { |
| 650 | if (SanityManager.DEBUG_ON(DUMP_LOG_ONLY)) |
| 651 | { |
| 652 | currentCheckpoint = null; |
| 653 | |
| 654 | System.out.println("Dump log only"); |
| 655 | |
| 656 | // unless otherwise specified, 1st log file starts at 1 |
| 657 | String beginLogFileNumber = |
| 658 | PropertyUtil.getSystemProperty( |
| 659 | DUMP_LOG_FROM_LOG_FILE); |
| 660 | |
| 661 | if (beginLogFileNumber != null) |
| 662 | { |
| 663 | logFileNumber = |
| 664 | Long.valueOf(beginLogFileNumber).longValue(); |
| 665 | } |
| 666 | else |
| 667 | { |
| 668 | logFileNumber = 1; |
| 669 | } |
| 670 | } |
| 671 | } |
| 672 | |
| 673 | if (SanityManager.DEBUG) |
| 674 | { |
| 675 | if (SanityManager.DEBUG_ON("setCheckpoint")) |
| 676 | { |
| 677 | currentCheckpoint = null; |
| 678 | |
| 679 | System.out.println("Set Checkpoint."); |
| 680 | |
| 681 | // unless otherwise specified, 1st log file starts at 1 |
| 682 | String checkpointStartLogStr = |
| 683 | PropertyUtil.getSystemProperty( |
| 684 | "derby.storage.checkpointStartLog"); |
| 685 | |
| 686 | String checkpointStartOffsetStr = |
| 687 | PropertyUtil.getSystemProperty( |
| 688 | "derby.storage.checkpointStartOffset"); |
| 689 | |
| 690 | |
| 691 | if ((checkpointStartLogStr != null) && |
| 692 | (checkpointStartOffsetStr != null)) |
| 693 | { |
| 694 | checkpointInstant = |
| 695 | LogCounter.makeLogInstantAsLong( |
| 696 | Long.valueOf(checkpointStartLogStr).longValue(), |
| 697 | Long.valueOf(checkpointStartOffsetStr).longValue()); |
| 698 | } |
| 699 | else |
| 700 | { |
| 701 | SanityManager.THROWASSERT( |
| 702 | "must set derby.storage.checkpointStartLog and derby.storage.checkpointStartOffset, if setting setCheckpoint."); |
| 703 | } |
| 704 | |
| 705 | currentCheckpoint = |
| 706 | findCheckpoint(checkpointInstant, logger); |
| 707 | } |
| 708 | } |
| 709 | |
| 710 | long redoLWM = LogCounter.INVALID_LOG_INSTANT; |
| 711 | long undoLWM = LogCounter.INVALID_LOG_INSTANT; |
| 712 | long ttabInstant = LogCounter.INVALID_LOG_INSTANT; |
| 713 | |
| 714 | StreamLogScan redoScan = null; |
| 715 | if (currentCheckpoint != null) |
| 716 | { |
| 717 | Formatable transactionTable = null; |
| 718 | |
| 719 | // RESOLVE: sku |
| 720 | // currentCheckpoint.getTransactionTable(); |
| 721 | |
| 722 | // need to set the transaction table before the undo |
| 723 | tf.useTransactionTable(transactionTable); |
| 724 | |
| 725 | redoLWM = currentCheckpoint.redoLWM(); |
| 726 | undoLWM = currentCheckpoint.undoLWM(); |
| 727 | |
| 728 | if (transactionTable != null) |
| 729 | ttabInstant = checkpointInstant; |
| 730 | |
| 731 | if (SanityManager.DEBUG) |
| 732 | { |
| 733 | if (SanityManager.DEBUG_ON(DBG_FLAG)) |
| 734 | { |
| 735 | SanityManager.DEBUG(DBG_FLAG, |
| 736 | "Found checkpoint at " + |
| 737 | LogCounter.toDebugString(checkpointInstant) + |
| 738 | " " + currentCheckpoint.toString()); |
| 739 | } |
| 740 | } |
| 741 | |
| 742 | firstLogFileNumber = LogCounter.getLogFileNumber(redoLWM); |
| 743 | |
| 744 | // figure out where the first interesting log file is. |
| 745 | if (LogCounter.getLogFileNumber(undoLWM) < |
| 746 | firstLogFileNumber) |
| 747 | { |
| 748 | firstLogFileNumber = |
| 749 | LogCounter.getLogFileNumber(undoLWM); |
| 750 | } |
| 751 | |
| 752 | |
| 753 | // if the checkpoint record doesn't have a transaction |
| 754 | // table, we need to rebuild it by scanning the log from |
| 755 | // the undoLWM. If it does have a transaction table, we |
| 756 | // only need to scan the log from the redoLWM |
| 757 | |
| 758 | redoScan = (StreamLogScan) |
| 759 | openForwardsScan(undoLWM, (LogInstant)null); |
| 760 | |
| 761 | } |
| 762 | else |
| 763 | { |
| 764 | // no checkpoint |
| 765 | tf.useTransactionTable((Formatable)null); |
| 766 | |
| 767 | long start = |
| 768 | LogCounter.makeLogInstantAsLong( |
| 769 | logFileNumber, LOG_FILE_HEADER_SIZE); |
| 770 | |
| 771 | // no checkpoint, start redo from the beginning of the |
| 772 | // file - assume this is the first log file |
| 773 | firstLogFileNumber = logFileNumber; |
| 774 | |
| 775 | redoScan = (StreamLogScan) |
| 776 | openForwardsScan(start, (LogInstant)null); |
| 777 | } |
| 778 | |
| 779 | // open a transaction that is used for redo and rollback |
| 780 | RawTransaction recoveryTransaction = |
| 781 | tf.startTransaction( |
| 782 | rsf, |
| 783 | ContextService.getFactory().getCurrentContextManager(), |
| 784 | AccessFactoryGlobals.USER_TRANS_NAME); |
| 785 | |
| 786 | // make this transaction aware that it is a recovery transaction |
| 787 | // and don't spew forth post commit work while replaying the log |
| 788 | recoveryTransaction.recoveryTransaction(); |
| 789 | |
| 790 | ///////////////////////////////////////////////////////////// |
| 791 | // |
| 792 | // Redo loop - in FileLogger |
| 793 | // |
| 794 | ///////////////////////////////////////////////////////////// |
| 795 | |
| 796 | // |
| 797 | // set log factory state to inRedo so that if redo caused any |
| 798 | // dirty page to be written from the cache, it won't flush the |
| 799 | // log since the end of the log has not been determined and we |
| 800 | // know the log record that caused the page to change has |
| 801 | // already been written to the log. We need the page write to |
| 802 | // go thru the log factory because if the redo has a problem, |
| 803 | // the log factory is corrupt and the only way we know not to |
| 804 | // write out the page in a checkpoint is if it check with the |
| 805 | // log factory, and that is done via a flush - we use the WAL |
| 806 | // protocol to stop corrupt pages from writing to the disk. |
| 807 | // |
| 808 | inRedo = true; |
| 809 | |
| 810 | long logEnd = |
| 811 | logger.redo( |
| 812 | recoveryTransaction, tf, redoScan, redoLWM, |
| 813 | ttabInstant); |
| 814 | |
| 815 | inRedo = false; |
| 816 | |
| 817 | |
| 818 | |
| 819 | // if we are only interested in dumping the log, don't alter |
| 820 | // the database and prevent anyone from using the log |
| 821 | if (SanityManager.DEBUG) |
| 822 | { |
| 823 | if (SanityManager.DEBUG_ON(LogToFile.DUMP_LOG_ONLY)) |
| 824 | { |
| 825 | Monitor.logMessage("_____________________________________________________"); |
| 826 | Monitor.logMessage("\n\t\t Log dump finished"); |
| 827 | Monitor.logMessage("_____________________________________________________"); |
| 828 | // just in case, it has not been set anyway |
| 829 | logOut = null; |
| 830 | |
| 831 | return; |
| 832 | } |
| 833 | } |
| 834 | |
| 835 | |
| 836 | ///////////////////////////////////////////////////////////// |
| 837 | // |
| 838 | // determine where the log ends |
| 839 | // |
| 840 | ///////////////////////////////////////////////////////////// |
| 841 | StorageRandomAccessFile theLog = null; |
| 842 | |
| 843 | |
| 844 | // if logend == LogCounter.INVALID_LOG_SCAN, that means there |
| 845 | // is no log record in the log - most likely it is corrupted in |
| 846 | // some way ... |
| 847 | if (logEnd == LogCounter.INVALID_LOG_INSTANT) |
| 848 | { |
| 849 | Monitor.logTextMessage(MessageId.LOG_LOG_NOT_FOUND); |
| 850 | |
| 851 | StorageFile logFile = getLogFileName(logFileNumber); |
| 852 | |
| 853 | if (privExists(logFile)) |
| 854 | { |
| 855 | // if we can delete this strange corrupted file, do so, |
| 856 | // otherwise, skip it |
| 857 | if (!privDelete(logFile)) |
| 858 | { |
| 859 | logFile = getLogFileName(++logFileNumber); |
| 860 | } |
| 861 | } |
| 862 | |
| 863 | try |
| 864 | { |
| 865 | theLog = privRandomAccessFile(logFile, "rw"); |
| 866 | } |
| 867 | catch (IOException ioe) |
| 868 | { |
| 869 | theLog = null; |
| 870 | } |
| 871 | |
| 872 | if (theLog == null || !privCanWrite(logFile)) |
| 873 | { |
| 874 | if (theLog != null) |
| 875 | theLog.close(); |
| 876 | |
| 877 | theLog = null; |
| 878 | |
| 879 | ReadOnlyDB = true; |
| 880 | } |
| 881 | else |
| 882 | { |
| 883 | try |
| 884 | { |
| 885 | // no previous log file or previous log position |
| 886 | if (!initLogFile( |
| 887 | theLog, logFileNumber, |
| 888 | LogCounter.INVALID_LOG_INSTANT)) |
| 889 | { |
| 890 | throw markCorrupt( |
| 891 | StandardException.newException( |
| 892 | SQLState.LOG_SEGMENT_NOT_EXIST, |
| 893 | logFile.getPath())); |
| 894 | } |
| 895 | } |
| 896 | catch (IOException ioe) |
| 897 | { |
| 898 | throw markCorrupt( |
| 899 | StandardException.newException( |
| 900 | SQLState.LOG_IO_ERROR, ioe)); |
| 901 | } |
| 902 | |
| 903 | // successfully init'd the log file - set up markers, |
| 904 | // and position at the end of the log. |
| 905 | endPosition = theLog.getFilePointer(); |
| 906 | lastFlush = endPosition; |
| 907 | |
| 908 | //if write sync is true , prellocate the log file |
| 909 | //and reopen the file in rws mode. |
| 910 | if(isWriteSynced) |
| 911 | { |
| 912 | //extend the file by wring zeros to it |
| 913 | preAllocateNewLogFile(theLog); |
| 914 | theLog.close(); |
| 915 | theLog = openLogFileInWriteMode(logFile); |
| 916 | //postion the log at the current end postion |
| 917 | theLog.seek(endPosition); |
| 918 | } |
| 919 | |
| 920 | if (SanityManager.DEBUG) |
| 921 | { |
| 922 | SanityManager.ASSERT( |
| 923 | endPosition == LOG_FILE_HEADER_SIZE, |
| 924 | "empty log file has wrong size"); |
| 925 | } |
| 926 | |
| 927 | //because we already incrementing the log number |
| 928 | //here, no special log switch required for |
| 929 | //backup recoveries. |
| 930 | logSwitchRequired = false; |
| 931 | } |
| 932 | } |
| 933 | else |
| 934 | { |
| 935 | // logEnd is the instant of the next log record in the log |
| 936 | // it is used to determine the last known good position of |
| 937 | // the log |
| 938 | logFileNumber = LogCounter.getLogFileNumber(logEnd); |
| 939 | |
| 940 | ReadOnlyDB = df.isReadOnly(); |
| 941 | |
| 942 | StorageFile logFile = getLogFileName(logFileNumber); |
| 943 | |
| 944 | if (!ReadOnlyDB) |
| 945 | { |
| 946 | // if datafactory doesn't think it is readonly, we can |
| 947 | // do some futher test of our own |
| 948 | try |
| 949 | { |
| 950 | if(isWriteSynced) |
| 951 | theLog = openLogFileInWriteMode(logFile); |
| 952 | else |
| 953 | theLog = privRandomAccessFile(logFile, "rw"); |
| 954 | } |
| 955 | catch (IOException ioe) |
| 956 | { |
| 957 | theLog = null; |
| 958 | } |
| 959 | if (theLog == null || !privCanWrite(logFile)) |
| 960 | { |
| 961 | if (theLog != null) |
| 962 | theLog.close(); |
| 963 | theLog = null; |
| 964 | |
| 965 | ReadOnlyDB = true; |
| 966 | } |
| 967 | } |
| 968 | |
| 969 | if (!ReadOnlyDB) |
| 970 | { |
| 971 | endPosition = LogCounter.getLogFilePosition(logEnd); |
| 972 | |
| 973 | // |
| 974 | // The end of the log is at endPosition. Which is where |
| 975 | // the next log should be appending. |
| 976 | // |
| 977 | // if the last log record ends before the end of the |
| 978 | // log file, then this log file has a fuzzy end. |
| 979 | // Zap all the bytes to between endPosition to EOF to 0. |
| 980 | // |
| 981 | // the end log marker is 4 bytes (of zeros) |
| 982 | // |
| 983 | // if endPosition + 4 == logOut.length, we have a |
| 984 | // properly terminated log file |
| 985 | // |
| 986 | // if endPosition + 4 is > logOut.length, there are 0, |
| 987 | // 1, 2, or 3 bytes of 'fuzz' at the end of the log. We |
| 988 | // can ignore that because it is guaranteed to be |
| 989 | // overwritten by the next log record. |
| 990 | // |
| 991 | // if endPosition + 4 is < logOut.length, we have a |
| 992 | // partial log record at the end of the log. |
| 993 | // |
| 994 | // We need to overwrite all of the incomplete log |
| 995 | // record, because if we start logging but cannot |
| 996 | // 'consume' all the bad log, then the log will truly |
| 997 | // be corrupted if the next 4 bytes (the length of the |
| 998 | // log record) after that is small enough that the next |
| 999 | // time the database is recovered, it will be |
| 1000 | // interpreted that the whole log record is in the log |
| 1001 | // and will try to objectify, only to get classNotFound |
| 1002 | // error or worse. |
| 1003 | // |
| 1004 | |
| 1005 | //find out if log had incomplete log records at the end. |
| 1006 | if (redoScan.isLogEndFuzzy()) |
| 1007 | { |
| 1008 | theLog.seek(endPosition); |
| 1009 | long eof = theLog.length(); |
| 1010 | |
| 1011 | Monitor.logTextMessage(MessageId.LOG_INCOMPLETE_LOG_RECORD, |
| 1012 | logFile, new Long(endPosition), new Long(eof)); |
| 1013 | |
| 1014 | /* Write zeros from incomplete log record to end of file */ |
| 1015 | long nWrites = (eof - endPosition)/logBufferSize; |
| 1016 | int rBytes = (int)((eof - endPosition) % logBufferSize); |
| 1017 | byte zeroBuf[]= new byte[logBufferSize]; |
| 1018 | |
| 1019 | //write the zeros to file |
| 1020 | while(nWrites-- > 0) |
| 1021 | theLog.write(zeroBuf); |
| 1022 | if(rBytes !=0) |
| 1023 | theLog.write(zeroBuf, 0, rBytes); |
| 1024 | |
| 1025 | if(!isWriteSynced) |
| 1026 | syncFile(theLog); |
| 1027 | } |
| 1028 | |
| 1029 | if (SanityManager.DEBUG) |
| 1030 | { |
| 1031 | if (theLog.length() != endPosition) |
| 1032 | { |
| 1033 | SanityManager.ASSERT( |
| 1034 | theLog.length() > endPosition, |
| 1035 | "log end > log file length, bad scan"); |
| 1036 | } |
| 1037 | } |
| 1038 | |
| 1039 | // set the log to the true end position, |
| 1040 | // and not the end of the file |
| 1041 | |
| 1042 | lastFlush = endPosition; |
| 1043 | theLog.seek(endPosition); |
| 1044 | } |
| 1045 | } |
| 1046 | |
| 1047 | if (theLog != null) |
| 1048 | logOut = new LogAccessFile(this, theLog, logBufferSize); |
| 1049 | |
| 1050 | if(logSwitchRequired) |
| 1051 | switchLogFile(); |
| 1052 | |
| 1053 | |
| 1054 | boolean noInFlightTransactions = tf.noActiveUpdateTransaction(); |
| 1055 | |
| 1056 | if (ReadOnlyDB) |
| 1057 | { |
| 1058 | // in the unlikely event that someone detects we are |
| 1059 | // dealing with a read only db, check to make sure the |
| 1060 | // database is quiesce when it was copied with no unflushed |
| 1061 | // dirty buffer |
| 1062 | if (!noInFlightTransactions) |
| 1063 | { |
| 1064 | throw StandardException.newException( |
| 1065 | SQLState.LOG_READ_ONLY_DB_NEEDS_UNDO); |
| 1066 | } |
| 1067 | } |
| 1068 | |
| 1069 | ///////////////////////////////////////////////////////////// |
| 1070 | // |
| 1071 | // Undo loop - in transaction factory. It just gets one |
| 1072 | // transaction at a time from the transaction table and calls |
| 1073 | // undo, no different from runtime. |
| 1074 | // |
| 1075 | ///////////////////////////////////////////////////////////// |
| 1076 | |
| 1077 | if (SanityManager.DEBUG) |
| 1078 | { |
| 1079 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 1080 | SanityManager.DEBUG(LogToFile.DBG_FLAG, |
| 1081 | "About to call undo(), transaction table =" + |
| 1082 | tf.getTransactionTable()); |
| 1083 | } |
| 1084 | |
| 1085 | if (!noInFlightTransactions) |
| 1086 | { |
| 1087 | if (SanityManager.DEBUG) |
| 1088 | { |
| 1089 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 1090 | SanityManager.DEBUG(LogToFile.DBG_FLAG, |
| 1091 | "In recovery undo, rollback inflight transactions"); |
| 1092 | } |
| 1093 | |
| 1094 | tf.rollbackAllTransactions(recoveryTransaction, rsf); |
| 1095 | |
| 1096 | if (SanityManager.DEBUG) |
| 1097 | { |
| 1098 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 1099 | SanityManager.DEBUG( |
| 1100 | LogToFile.DBG_FLAG, "finish recovery undo,"); |
| 1101 | } |
| 1102 | } |
| 1103 | else |
| 1104 | { |
| 1105 | if (SanityManager.DEBUG) |
| 1106 | { |
| 1107 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 1108 | SanityManager.DEBUG(LogToFile.DBG_FLAG, |
| 1109 | "No in flight transaction, no recovery undo work"); |
| 1110 | } |
| 1111 | } |
| 1112 | |
| 1113 | ///////////////////////////////////////////////////////////// |
| 1114 | // |
| 1115 | // XA prepared xact loop - in transaction factory. At this |
| 1116 | // point only prepared transactions should be left in the |
| 1117 | // transaction table, all others should have been aborted or |
| 1118 | // committed and removed from the transaction table. It just |
| 1119 | // gets one transaction at a time from the transaction table, |
| 1120 | // creates a real context and transaction, reclaims locks, |
| 1121 | // and leaves the new xact in the transaction table. |
| 1122 | // |
| 1123 | ///////////////////////////////////////////////////////////// |
| 1124 | |
| 1125 | if (SanityManager.DEBUG) |
| 1126 | { |
| 1127 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 1128 | SanityManager.DEBUG(LogToFile.DBG_FLAG, |
| 1129 | "About to call rePrepare(), transaction table =" + |
| 1130 | tf.getTransactionTable()); |
| 1131 | } |
| 1132 | |
| 1133 | tf.handlePreparedXacts(rsf); |
| 1134 | |
| 1135 | if (SanityManager.DEBUG) |
| 1136 | { |
| 1137 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 1138 | SanityManager.DEBUG(LogToFile.DBG_FLAG, |
| 1139 | "Finished rePrepare(), transaction table =" + |
| 1140 | tf.getTransactionTable()); |
| 1141 | } |
| 1142 | |
| 1143 | ///////////////////////////////////////////////////////////// |
| 1144 | // |
| 1145 | // End of recovery. |
| 1146 | // |
| 1147 | ///////////////////////////////////////////////////////////// |
| 1148 | |
| 1149 | // recovery is finished. Close the transaction |
| 1150 | recoveryTransaction.close(); |
| 1151 | |
| 1152 | |
| 1153 | // notify the dataFactory that recovery is completed, |
| 1154 | // but before the checkpoint is written. |
| 1155 | dataFactory.postRecovery(); |
| 1156 | |
| 1157 | |
| 1158 | ////////////////////////////////////////////////////////////// |
| 1159 | // set the transaction factory short id, we have seen all the |
| 1160 | // trasactions in the log, and at the minimum, the checkpoint |
| 1161 | // transaction will be there. Set the shortId to the next |
| 1162 | // value. |
| 1163 | ////////////////////////////////////////////////////////////// |
| 1164 | tf.resetTranId(); |
| 1165 | |
| 1166 | // do a checkpoint (will flush the log) if there is any rollback |
| 1167 | // if can't checkpoint for some reasons, flush log and carry on |
| 1168 | if (!ReadOnlyDB) |
| 1169 | { |
| 1170 | boolean needCheckpoint = true; |
| 1171 | |
| 1172 | // if we can figure out there there is very little in the |
| 1173 | // log (less than 1000 bytes), we haven't done any |
| 1174 | // rollbacks, then don't checkpoint. Otherwise checkpoint. |
| 1175 | if (currentCheckpoint != null && noInFlightTransactions && |
| 1176 | redoLWM != LogCounter.INVALID_LOG_INSTANT && |
| 1177 | undoLWM != LogCounter.INVALID_LOG_INSTANT) |
| 1178 | { |
| 1179 | if ((logFileNumber == LogCounter.getLogFileNumber(redoLWM)) |
| 1180 | && (logFileNumber == LogCounter.getLogFileNumber(undoLWM)) |
| 1181 | && (endPosition < (LogCounter.getLogFilePosition(redoLWM) + 1000))) |
| 1182 | needCheckpoint = false; |
| 1183 | } |
| 1184 | |
| 1185 | if (needCheckpoint && !checkpoint(rsf, df, tf, false)) |
| 1186 | flush(logFileNumber, endPosition); |
| 1187 | } |
| 1188 | |
| 1189 | logger.close(); |
| 1190 | |
| 1191 | recoveryNeeded = false; |
| 1192 | } |
| 1193 | catch (IOException ioe) |
| 1194 | { |
| 1195 | if (SanityManager.DEBUG) |
| 1196 | ioe.printStackTrace(); |
| 1197 | |
| 1198 | throw markCorrupt( |
| 1199 | StandardException.newException(SQLState.LOG_IO_ERROR, ioe)); |
| 1200 | } |
| 1201 | catch (ClassNotFoundException cnfe) |
| 1202 | { |
| 1203 | throw markCorrupt( |
| 1204 | StandardException.newException( |
| 1205 | SQLState.LOG_CORRUPTED, cnfe)); |
| 1206 | } |
| 1207 | catch (StandardException se) |
| 1208 | { |
| 1209 | throw markCorrupt(se); |
| 1210 | } |
| 1211 | catch (Throwable th) |
| 1212 | { |
| 1213 | if (SanityManager.DEBUG) |
| 1214 | { |
| 1215 | SanityManager.showTrace(th); |
| 1216 | th.printStackTrace(); |
| 1217 | } |
| 1218 | |
| 1219 | throw markCorrupt( |
| 1220 | StandardException.newException( |
| 1221 | SQLState.LOG_RECOVERY_FAILED, th)); |
| 1222 | } |
| 1223 | } |
| 1224 | else |
| 1225 | { |
| 1226 | |
| 1227 | tf.useTransactionTable((Formatable)null); |
| 1228 | |
| 1229 | // set the transaction factory short id |
| 1230 | tf.resetTranId(); |
| 1231 | } |
| 1232 | |
| 1233 | // done with recovery |
| 1234 | |
| 1235 | ///////////////////////////////////////////////////////////// |
| 1236 | // setup checktpoint daemon |
| 1237 | ///////////////////////////////////////////////////////////// |
| 1238 | checkpointDaemon = rawStoreFactory.getDaemon(); |
| 1239 | if (checkpointDaemon != null) |
| 1240 | { |
| 1241 | myClientNumber = |
| 1242 | checkpointDaemon.subscribe(this, true /*onDemandOnly */); |
| 1243 | } |
| 1244 | } |
| 1245 | |
| 1246 | |
| 1247 | /** |
| 1248 | Checkpoint the rawStore. |
| 1249 | |
| 1250 | <P> MT- Only one checkpoint is to be taking place at any given time. |
| 1251 | |
| 1252 | <P> The steps of a checkpoint are |
| 1253 | |
| 1254 | <OL> |
| 1255 | <LI> switch to a new log file if possible |
| 1256 | <PRE> |
| 1257 | freeze the log (for the transition to a new log file) |
| 1258 | flush current log file |
| 1259 | create and flush the new log file (with file number 1 higher |
| 1260 | than the previous log file). The new log file becomes the |
| 1261 | current log file. |
| 1262 | unfreeze the log |
| 1263 | </PRE> |
| 1264 | <LI> start checkpoint transaction |
| 1265 | <LI> gather interesting information about the rawStore: |
| 1266 | the current log instant (redoLWM) |
| 1267 | the earliest active transaction begin tran log record |
| 1268 | instant (undoLWM), all the truncation LWM set by clients |
| 1269 | of raw store (replication) |
| 1270 | <LI> clean the buffer cache |
| 1271 | <LI> log the next checkpoint log record, which contains |
| 1272 | (repPoint, undoLWM, redoLWM) and commit checkpoint transaction. |
| 1273 | <LI> synchronously write the control file containing the next checkpoint |
| 1274 | log record log instant |
| 1275 | <LI> the new checkpoint becomes the current checkpoint. |
| 1276 | Somewhere near the beginning of each log file should be a |
| 1277 | checkpoint log record (not guarenteed to be there) |
| 1278 | <LI> see if the log can be truncated |
| 1279 | |
| 1280 | <P> |
| 1281 | The earliest useful log record is determined by the repPoint and the |
| 1282 | undoLWM, whichever is earlier. |
| 1283 | <P> |
| 1284 | Every log file whose log file number is smaller than the earliest |
| 1285 | useful log record's log file number can be deleted. |
| 1286 | |
| 1287 | <P><PRE> |
| 1288 | Transactions can be at the following states w/r to a checkpoint - |
| 1289 | consider the log as a continous stream and not as series of log |
| 1290 | files for the sake of clarity. |
| 1291 | |(BT)-------(ET)| marks the begin and end of a transaction. |
| 1292 | . checkpoint started |
| 1293 | . |__undoLWM | |
| 1294 | . V |___redoLWM |
| 1295 | . |___TruncationLWM |
| 1296 | . | |
| 1297 | . V |
| 1298 | 1 |-----------------| |
| 1299 | 2 |--------------------------------| |
| 1300 | 3 |-------| |
| 1301 | 4 |--------------------------------------(end of log) |
| 1302 | 5 |-^-| |
| 1303 | . Checkpoint Log Record |
| 1304 | ---A--->|<-------B--------->|<-------------C----------- |
| 1305 | </PRE> |
| 1306 | |
| 1307 | <P> |
| 1308 | There are only 3 periods of interest : <BR> |
| 1309 | A) before undoLWM, B) between undo and redo LWM, C) after redoLWM. |
| 1310 | |
| 1311 | <P> |
| 1312 | Transaction 1 started in A and terminates in B.<BR> |
| 1313 | During redo, we should only see log records and endXact from this |
| 1314 | transaction in the first phase (between undoLWM and redoLWM). No |
| 1315 | beginXact log record for this transaction will be seen. |
| 1316 | |
| 1317 | <P> |
| 1318 | Transaction 2 started in B (right on the undoLWM) and terminated in C.<BR> |
| 1319 | Any transaction that terminates in C must have a beginXact at or |
| 1320 | after undoLWM. In other words, no transaction can span A, B and C. |
| 1321 | During redo, we will see beginXact, other log records and endXact |
| 1322 | for this transaction. |
| 1323 | |
| 1324 | <P> |
| 1325 | Transaction 3 started in B and ended in B.<BR> |
| 1326 | During redo, we will see beginXact, other log records and endXact |
| 1327 | for this transaction. |
| 1328 | |
| 1329 | <P> |
| 1330 | Transaction 4 begins in B and never ends.<BR> |
| 1331 | During redo, we will see beginXact, other log records. |
| 1332 | In undo, this loser transaction will be rolled back. |
| 1333 | |
| 1334 | <P> |
| 1335 | Transaction 5 is the transaction taking the checkpoint.<BR> |
| 1336 | The checkpoint action started way back in time but the checkpoint |
| 1337 | log record is only written after the buffer cache has been flushed. |
| 1338 | |
| 1339 | <P> |
| 1340 | Note that if any time elapse between taking the undoLWM and the |
| 1341 | redoLWM, then it will create a 4th period of interest. |
| 1342 | |
| 1343 | @exception StandardException - encounter exception while doing |
| 1344 | checkpoint. |
| 1345 | */ |
| 1346 | public boolean checkpoint(RawStoreFactory rsf, |
| 1347 | DataFactory df, |
| 1348 | TransactionFactory tf, |
| 1349 | boolean wait) |
| 1350 | throws StandardException |
| 1351 | { |
| 1352 | |
| 1353 | // call checkpoint with no pre-started transaction |
| 1354 | boolean done = checkpointWithTran(null, rsf, df, tf); |
| 1355 | |
| 1356 | //above checpoint call will return 'false' without |
| 1357 | //performing the checkpoint if some other thread is doing checkpoint. |
| 1358 | //In cases like backup it is necesary to wait for the |
| 1359 | //checkpoint to complete before copying the files. 'wait' flag get passed |
| 1360 | //in as 'true' by such cases. |
| 1361 | //When wait flag is true, we will wait here until the other thread which |
| 1362 | //is actually doing the the checkpoint completes. |
| 1363 | |
| 1364 | if(!done && wait) |
| 1365 | { |
| 1366 | synchronized(this) |
| 1367 | { |
| 1368 | //wait until the thread that is doing the checkpoint completes it. |
| 1369 | while(inCheckpoint) |
| 1370 | { |
| 1371 | try |
| 1372 | { |
| 1373 | wait(); |
| 1374 | } |
| 1375 | catch (InterruptedException ie) |
| 1376 | { |
| 1377 | throw StandardException.interrupt(ie); |
| 1378 | } |
| 1379 | } |
| 1380 | done = true; |
| 1381 | } |
| 1382 | } |
| 1383 | |
| 1384 | return done; |
| 1385 | } |
| 1386 | |
| 1387 | |
| 1388 | /** |
| 1389 | checkpoint with pre-start transaction |
| 1390 | |
| 1391 | @exception StandardException Cloudscape Standard Error Policy |
| 1392 | */ |
| 1393 | protected boolean checkpointWithTran(RawTransaction cptran, |
| 1394 | RawStoreFactory rsf, |
| 1395 | DataFactory df, |
| 1396 | TransactionFactory tf) |
| 1397 | throws StandardException |
| 1398 | { |
| 1399 | boolean proceed = true; |
| 1400 | LogInstant redoLWM; |
| 1401 | |
| 1402 | // we may be called to stop the database after a bad error, make sure |
| 1403 | // logout is set |
| 1404 | if (logOut == null) |
| 1405 | { |
| 1406 | return false; |
| 1407 | } |
| 1408 | |
| 1409 | long approxLogLength; |
| 1410 | |
| 1411 | synchronized (this) |
| 1412 | { |
| 1413 | // has someone else found a problem in the raw store? |
| 1414 | if (corrupt != null) |
| 1415 | { |
| 1416 | throw StandardException.newException(SQLState.LOG_STORE_CORRUPT, corrupt); |
| 1417 | } |
| 1418 | |
| 1419 | // if another checkpoint is in progress, don't do anything |
| 1420 | if (inCheckpoint == true) |
| 1421 | proceed = false; |
| 1422 | else |
| 1423 | inCheckpoint = true; |
| 1424 | |
| 1425 | approxLogLength = endPosition; // current end position |
| 1426 | |
| 1427 | // don't return from inside of a sync block |
| 1428 | } |
| 1429 | |
| 1430 | if (!proceed) |
| 1431 | { |
| 1432 | return false; |
| 1433 | } |
| 1434 | |
| 1435 | // needCPtran == true if we are not supplied with a pre-started transaction |
| 1436 | boolean needCPTran = (cptran == null); |
| 1437 | |
| 1438 | if (SanityManager.DEBUG) |
| 1439 | { |
| 1440 | if (logSwitchInterval == 0) |
| 1441 | { |
| 1442 | SanityManager.THROWASSERT( |
| 1443 | "switching log file: Approx log length = " + |
| 1444 | approxLogLength + " logSwitchInterval = 0"); |
| 1445 | } |
| 1446 | } |
| 1447 | |
| 1448 | |
| 1449 | try |
| 1450 | { |
| 1451 | if (approxLogLength > logSwitchInterval) |
| 1452 | { |
| 1453 | switchLogFile(); |
| 1454 | //log switch is occuring in conjuction with the |
| 1455 | //checkpoint, set the amount of log written from last checkpoint to zero. |
| 1456 | logWrittenFromLastCheckPoint = 0; |
| 1457 | }else |
| 1458 | { |
| 1459 | //checkpoint is happening without the log switch, |
| 1460 | //in the middle of a log file. Amount of log written already for |
| 1461 | //the current log file should not be included in caluculation |
| 1462 | //of when next check point is due. By assigning the negative |
| 1463 | //value of amount of log writtent for this file. Later it will |
| 1464 | //be subtracted when we switch the log file or while calculating whether |
| 1465 | //we are due a for checkpoint a flush time. |
| 1466 | logWrittenFromLastCheckPoint = -endPosition; |
| 1467 | } |
| 1468 | |
| 1469 | if (SanityManager.DEBUG) |
| 1470 | { |
| 1471 | // if this debug flag is set on, just switch log file |
| 1472 | |
| 1473 | if (SanityManager.DEBUG_ON(TEST_LOG_SWITCH_LOG)) |
| 1474 | return false; |
| 1475 | } |
| 1476 | |
| 1477 | |
| 1478 | // start a checkpoint transaction |
| 1479 | if (needCPTran) |
| 1480 | cptran = tf.startInternalTransaction(rsf, |
| 1481 | ContextService.getFactory().getCurrentContextManager()); |
| 1482 | |
| 1483 | ///////////////////////////////////////////////////// |
| 1484 | // gather a snapshot of the various interesting points of the log |
| 1485 | ///////////////////////////////////////////////////// |
| 1486 | long undoLWM_long; |
| 1487 | long redoLWM_long; |
| 1488 | |
| 1489 | synchronized(this) // we could synchronized on something else, it |
| 1490 | // doesn't matter as long as logAndDo sync on |
| 1491 | // the same thing |
| 1492 | { |
| 1493 | // The redo LWM is the current log instant. We are going to |
| 1494 | // clean the cache shortly, any log record before this point |
| 1495 | // will not ever need to be redone. |
| 1496 | redoLWM_long = currentInstant(); |
| 1497 | redoLWM = new LogCounter(redoLWM_long); |
| 1498 | |
| 1499 | // The undo LWM is what we need to rollback all transactions. |
| 1500 | // Synchronize this with the starting of a new transaction so |
| 1501 | // that the transaction factory can have a consistent view |
| 1502 | // See FileLogger.logAndDo |
| 1503 | |
| 1504 | LogCounter undoLWM = (LogCounter)(tf.firstUpdateInstant()); |
| 1505 | if (undoLWM == null) |
| 1506 | undoLWM_long = redoLWM_long; // no active transaction |
| 1507 | else |
| 1508 | undoLWM_long = undoLWM.getValueAsLong(); |
| 1509 | |
| 1510 | } |
| 1511 | |
| 1512 | ///////////////////////////////////////////////////// |
| 1513 | // clean the buffer cache |
| 1514 | ///////////////////////////////////////////////////// |
| 1515 | df.checkpoint(); |
| 1516 | |
| 1517 | |
| 1518 | ///////////////////////////////////////////////////// |
| 1519 | // write out the checkpoint log record |
| 1520 | ///////////////////////////////////////////////////// |
| 1521 | |
| 1522 | // send the checkpoint record to the log |
| 1523 | Formatable transactionTable = tf.getTransactionTable(); |
| 1524 | |
| 1525 | CheckpointOperation nextCheckpoint = |
| 1526 | new CheckpointOperation( |
| 1527 | redoLWM_long, undoLWM_long, transactionTable); |
| 1528 | |
| 1529 | cptran.logAndDo(nextCheckpoint); |
| 1530 | |
| 1531 | LogCounter checkpointInstant = |
| 1532 | (LogCounter)(cptran.getLastLogInstant()); |
| 1533 | |
| 1534 | if (checkpointInstant != null) |
| 1535 | { |
| 1536 | // since checkpoint is an internal transaction, I need to |
| 1537 | // flush it to make sure it actually goes to the log |
| 1538 | flush(checkpointInstant); |
| 1539 | } |
| 1540 | else |
| 1541 | { |
| 1542 | throw StandardException.newException( |
| 1543 | SQLState.LOG_CANNOT_LOG_CHECKPOINT); |
| 1544 | } |
| 1545 | |
| 1546 | cptran.commit(); |
| 1547 | |
| 1548 | if (needCPTran) |
| 1549 | { |
| 1550 | cptran.close(); // if we started it, we will close it |
| 1551 | cptran = null; |
| 1552 | } |
| 1553 | |
| 1554 | ///////////////////////////////////////////////////// |
| 1555 | // write out the log control file which contains the last |
| 1556 | // successful checkpoint log record |
| 1557 | ///////////////////////////////////////////////////// |
| 1558 | |
| 1559 | if (!writeControlFile(getControlFileName(), |
| 1560 | checkpointInstant.getValueAsLong())) |
| 1561 | { |
| 1562 | throw StandardException.newException( |
| 1563 | SQLState.LOG_CONTROL_FILE, getControlFileName()); |
| 1564 | } |
| 1565 | |
| 1566 | // next checkpoint becomes the current checkpoint |
| 1567 | currentCheckpoint = nextCheckpoint; |
| 1568 | |
| 1569 | |
| 1570 | //////////////////////////////////////////////////// |
| 1571 | // see if we can reclaim some log space |
| 1572 | //////////////////////////////////////////////////// |
| 1573 | |
| 1574 | if (!logArchived()) |
| 1575 | { |
| 1576 | truncateLog(currentCheckpoint); |
| 1577 | } |
| 1578 | |
| 1579 | // delete the committted container drop stubs |
| 1580 | // that are no longer required during recovery. |
| 1581 | // If a backup is in progress don't delete the stubs until |
| 1582 | // it is done. Backup needs to copy all the stubs that |
| 1583 | // are needed to recover from the backup checkpoint on restore. |
| 1584 | if(!backupInProgress) |
| 1585 | df.removeDroppedContainerFileStubs(redoLWM); |
| 1586 | |
| 1587 | } |
| 1588 | catch (IOException ioe) |
| 1589 | { |
| 1590 | throw markCorrupt( |
| 1591 | StandardException.newException(SQLState.LOG_IO_ERROR, ioe)); |
| 1592 | } |
| 1593 | finally |
| 1594 | { |
| 1595 | synchronized(this) |
| 1596 | { |
| 1597 | |
| 1598 | |
| 1599 | inCheckpoint = false; |
| 1600 | notifyAll(); |
| 1601 | } |
| 1602 | |
| 1603 | if (cptran != null && needCPTran) |
| 1604 | { |
| 1605 | try |
| 1606 | { |
| 1607 | cptran.commit(); |
| 1608 | cptran.close(); |
| 1609 | } |
| 1610 | catch (StandardException se) |
| 1611 | { |
| 1612 | throw markCorrupt(StandardException.newException( |
| 1613 | SQLState.LOG_CORRUPTED, se)); |
| 1614 | } |
| 1615 | } |
| 1616 | } |
| 1617 | |
| 1618 | return true; |
| 1619 | } |
| 1620 | |
| 1621 | /** |
| 1622 | Flush all unwritten log record up to the log instance indicated to disk |
| 1623 | and sync. |
| 1624 | Also check to see if database is frozen or corrupt. |
| 1625 | |
| 1626 | <P>MT - not needed, wrapper method |
| 1627 | |
| 1628 | @param where flush log up to here |
| 1629 | |
| 1630 | @exception StandardException Standard Cloudscape error policy |
| 1631 | */ |
| 1632 | public void flush(LogInstant where) throws StandardException |
| 1633 | { |
| 1634 | long fileNumber; |
| 1635 | long wherePosition; |
| 1636 | |
| 1637 | if (where == null) { |
| 1638 | // don't flush, just use this to check if database is frozen or |
| 1639 | // corrupt |
| 1640 | fileNumber = 0; |
| 1641 | wherePosition = LogCounter.INVALID_LOG_INSTANT; |
| 1642 | } else { |
| 1643 | LogCounter whereC = (LogCounter) where; |
| 1644 | fileNumber = whereC.getLogFileNumber(); |
| 1645 | wherePosition = whereC.getLogFilePosition(); |
| 1646 | } |
| 1647 | flush(fileNumber, wherePosition); |
| 1648 | } |
| 1649 | |
| 1650 | /** |
| 1651 | Flush all unwritten log record to disk and sync. |
| 1652 | Also check to see if database is frozen or corrupt. |
| 1653 | |
| 1654 | <P>MT - not needed, wrapper method |
| 1655 | |
| 1656 | @exception StandardException Standard Cloudscape error policy |
| 1657 | */ |
| 1658 | public void flushAll() throws StandardException |
| 1659 | { |
| 1660 | long fnum; |
| 1661 | long whereTo; |
| 1662 | |
| 1663 | synchronized(this) |
| 1664 | { |
| 1665 | fnum = logFileNumber; |
| 1666 | whereTo = endPosition; |
| 1667 | } |
| 1668 | |
| 1669 | flush(fnum, whereTo); |
| 1670 | } |
| 1671 | |
| 1672 | /* |
| 1673 | * Private methods that helps to implement methods of LogFactory |
| 1674 | */ |
| 1675 | |
| 1676 | /** |
| 1677 | Verify that we the log file is of the right format and of the right |
| 1678 | version and log file number. |
| 1679 | |
| 1680 | <P>MT - not needed, no global variables used |
| 1681 | |
| 1682 | @param logFileName the name of the log file |
| 1683 | @param number the log file number |
| 1684 | @return true if the log file is of the current version and of the |
| 1685 | correct format |
| 1686 | |
| 1687 | @exception StandardException Standard Cloudscape error policy |
| 1688 | */ |
| 1689 | private boolean verifyLogFormat(StorageFile logFileName, long number) |
| 1690 | throws StandardException |
| 1691 | { |
| 1692 | boolean ret = false; |
| 1693 | try |
| 1694 | { |
| 1695 | StorageRandomAccessFile log = privRandomAccessFile(logFileName, "r"); |
| 1696 | ret = verifyLogFormat(log, number); |
| 1697 | log.close(); |
| 1698 | } |
| 1699 | catch (IOException ioe) |
| 1700 | { |
| 1701 | |
| 1702 | } |
| 1703 | |
| 1704 | return ret; |
| 1705 | } |
| 1706 | |
| 1707 | /** |
| 1708 | Verify that we the log file is of the right format and of the right |
| 1709 | version and log file number. The log file position is set to the |
| 1710 | beginning. |
| 1711 | |
| 1712 | <P>MT - MT-unsafe, caller must synchronize |
| 1713 | |
| 1714 | @param log the log file |
| 1715 | @param number the log file number |
| 1716 | @return true if the log file is of the current version and of the |
| 1717 | correct format |
| 1718 | |
| 1719 | @exception StandardException Standard Cloudscape error policy |
| 1720 | */ |
| 1721 | private boolean verifyLogFormat(StorageRandomAccessFile log, long number) |
| 1722 | throws StandardException |
| 1723 | { |
| 1724 | try |
| 1725 | { |
| 1726 | log.seek(0); |
| 1727 | int logfid = log.readInt(); |
| 1728 | int obsoleteLogVersion = log.readInt(); // this value is useless, for |
| 1729 | // backwards compatibility |
| 1730 | long logNumber = log.readLong(); |
| 1731 | |
| 1732 | if (logfid != fid || logNumber != number) |
| 1733 | { |
| 1734 | throw StandardException.newException( |
| 1735 | SQLState.LOG_INCOMPATIBLE_FORMAT, dataDirectory); |
| 1736 | } |
| 1737 | } |
| 1738 | catch (IOException ioe) |
| 1739 | { |
| 1740 | throw StandardException.newException( |
| 1741 | SQLState.LOG_CANNOT_VERIFY_LOG_FORMAT, ioe, dataDirectory); |
| 1742 | } |
| 1743 | |
| 1744 | return true; |
| 1745 | } |
| 1746 | |
| 1747 | /** |
| 1748 | Initialize the log to the correct format with the given version and |
| 1749 | log file number. The new log file must be empty. After initializing, |
| 1750 | the file is synchronously written to disk. |
| 1751 | |
| 1752 | <P>MT - synchornization provided by caller |
| 1753 | |
| 1754 | @param newlog the new log file to be initialized |
| 1755 | @param number the log file number |
| 1756 | @param prevLogRecordEndInstant the end position of the previous log record |
| 1757 | |
| 1758 | @return true if the log file is empty, else false. |
| 1759 | |
| 1760 | @exception IOException if new log file cannot be accessed or initialized |
| 1761 | */ |
| 1762 | |
| 1763 | private boolean initLogFile(StorageRandomAccessFile newlog, long number, |
| 1764 | long prevLogRecordEndInstant) |
| 1765 | throws IOException, StandardException |
| 1766 | { |
| 1767 | if (newlog.length() != 0) |
| 1768 | return false; |
| 1769 | |
| 1770 | if (SanityManager.DEBUG) |
| 1771 | { |
| 1772 | if ( SanityManager.DEBUG_ON(TEST_LOG_FULL)) |
| 1773 | testLogFull(); |
| 1774 | } |
| 1775 | if (SanityManager.DEBUG) |
| 1776 | { |
| 1777 | if (SanityManager.DEBUG_ON(TEST_SWITCH_LOG_FAIL1)) |
| 1778 | throw new IOException("TestLogSwitchFail1"); |
| 1779 | } |
| 1780 | |
| 1781 | |
| 1782 | newlog.seek(0); |
| 1783 | |
| 1784 | newlog.writeInt(fid); |
| 1785 | newlog.writeInt(OBSOLETE_LOG_VERSION_NUMBER); // for silly backwards compatibility reason |
| 1786 | newlog.writeLong(number); |
| 1787 | newlog.writeLong(prevLogRecordEndInstant); |
| 1788 | |
| 1789 | syncFile(newlog); |
| 1790 | |
| 1791 | return true; |
| 1792 | } |
| 1793 | |
| 1794 | /** |
| 1795 | Switch to the next log file if possible. |
| 1796 | |
| 1797 | <P>MT - log factory is single threaded thru a log file switch, the log |
| 1798 | is frozen for the duration of the switch |
| 1799 | */ |
| 1800 | private void switchLogFile() throws StandardException |
| 1801 | { |
| 1802 | boolean switchedOver = false; |
| 1803 | |
| 1804 | ///////////////////////////////////////////////////// |
| 1805 | // Freeze the log for the switch over to a new log file. |
| 1806 | // This blocks out any other threads from sending log |
| 1807 | // record to the log stream. |
| 1808 | // |
| 1809 | // The switching of the log file and checkpoint are really |
| 1810 | // independent events, they are tied together just because a |
| 1811 | // checkpoint is the natural place to switch the log and vice |
| 1812 | // versa. This could happen before the cache is flushed or |
| 1813 | // after the checkpoint log record is written. |
| 1814 | ///////////////////////////////////////////////////// |
| 1815 | synchronized (this) |
| 1816 | { |
| 1817 | |
| 1818 | // Make sure that this thread of control is guaranteed to complete |
| 1819 | // it's work of switching the log file without having to give up |
| 1820 | // the semaphore to a backup or another flusher. Do this by looping |
| 1821 | // until we have the semaphore, the log is not being flushed, and |
| 1822 | // the log is not frozen for backup. Track (2985). |
| 1823 | while(logBeingFlushed | isFrozen) |
| 1824 | { |
| 1825 | try |
| 1826 | { |
| 1827 | wait(); |
| 1828 | } |
| 1829 | catch (InterruptedException ie) |
| 1830 | { |
| 1831 | throw StandardException.interrupt(ie); |
| 1832 | } |
| 1833 | } |
| 1834 | |
| 1835 | // we have an empty log file here, refuse to switch. |
| 1836 | if (endPosition == LOG_FILE_HEADER_SIZE) |
| 1837 | { |
| 1838 | if (SanityManager.DEBUG) |
| 1839 | { |
| 1840 | Monitor.logMessage("not switching from an empty log file (" + |
| 1841 | logFileNumber + ")"); |
| 1842 | } |
| 1843 | return; |
| 1844 | } |
| 1845 | |
| 1846 | // log file isn't being flushed right now and logOut is not being |
| 1847 | // used. |
| 1848 | StorageFile newLogFile = getLogFileName(logFileNumber+1); |
| 1849 | |
| 1850 | if (logFileNumber+1 >= maxLogFileNumber) |
| 1851 | { |
| 1852 | throw StandardException.newException( |
| 1853 | SQLState.LOG_EXCEED_MAX_LOG_FILE_NUMBER, |
| 1854 | new Long(maxLogFileNumber)); |
| 1855 | } |
| 1856 | |
| 1857 | StorageRandomAccessFile newLog = null; // the new log file |
| 1858 | try |
| 1859 | { |
| 1860 | // if the log file exist and cannot be deleted, cannot |
| 1861 | // switch log right now |
| 1862 | if (privExists(newLogFile) && !privDelete(newLogFile)) |
| 1863 | { |
| 1864 | logErrMsg(MessageService.getTextMessage( |
| 1865 | MessageId.LOG_NEW_LOGFILE_EXIST, |
| 1866 | newLogFile.getPath())); |
| 1867 | return; |
| 1868 | } |
| 1869 | |
| 1870 | try |
| 1871 | { |
| 1872 | newLog = privRandomAccessFile(newLogFile, "rw"); |
| 1873 | } |
| 1874 | catch (IOException ioe) |
| 1875 | { |
| 1876 | newLog = null; |
| 1877 | } |
| 1878 | |
| 1879 | if (newLog == null || !privCanWrite(newLogFile)) |
| 1880 | { |
| 1881 | if (newLog != null) |
| 1882 | newLog.close(); |
| 1883 | newLog = null; |
| 1884 | |
| 1885 | return; |
| 1886 | } |
| 1887 | |
| 1888 | if (initLogFile(newLog, logFileNumber+1, |
| 1889 | LogCounter.makeLogInstantAsLong(logFileNumber, endPosition))) |
| 1890 | { |
| 1891 | |
| 1892 | // New log file init ok, close the old one and |
| 1893 | // switch over, after this point, need to shutdown the |
| 1894 | // database if any error crops up |
| 1895 | switchedOver = true; |
| 1896 | |
| 1897 | // write out an extra 0 at the end to mark the end of the log |
| 1898 | // file. |
| 1899 | |
| 1900 | logOut.writeEndMarker(0); |
| 1901 | |
| 1902 | endPosition += 4; |
| 1903 | //set that we are in log switch to prevent flusher |
| 1904 | //not requesting to switch log again |
| 1905 | inLogSwitch = true; |
| 1906 | // flush everything including the int we just wrote |
| 1907 | flush(logFileNumber, endPosition); |
| 1908 | |
| 1909 | |
| 1910 | // simulate out of log error after the switch over |
| 1911 | if (SanityManager.DEBUG) |
| 1912 | { |
| 1913 | if (SanityManager.DEBUG_ON(TEST_SWITCH_LOG_FAIL2)) |
| 1914 | throw new IOException("TestLogSwitchFail2"); |
| 1915 | } |
| 1916 | |
| 1917 | |
| 1918 | logOut.close(); // close the old log file |
| 1919 | |
| 1920 | logWrittenFromLastCheckPoint += endPosition; |
| 1921 | |
| 1922 | endPosition = newLog.getFilePointer(); |
| 1923 | lastFlush = endPosition; |
| 1924 | |
| 1925 | if(isWriteSynced) |
| 1926 | { |
| 1927 | //extend the file by wring zeros to it |
| 1928 | preAllocateNewLogFile(newLog); |
| 1929 | newLog.close(); |
| 1930 | newLog = openLogFileInWriteMode(newLogFile); |
| 1931 | newLog.seek(endPosition); |
| 1932 | } |
| 1933 | |
| 1934 | logOut = new LogAccessFile(this, newLog, logBufferSize); |
| 1935 | newLog = null; |
| 1936 | |
| 1937 | |
| 1938 | if (SanityManager.DEBUG) |
| 1939 | { |
| 1940 | if (endPosition != LOG_FILE_HEADER_SIZE) |
| 1941 | SanityManager.THROWASSERT( |
| 1942 | "new log file has unexpected size" + |
| 1943 | + endPosition); |
| 1944 | } |
| 1945 | logFileNumber++; |
| 1946 | |
| 1947 | if (SanityManager.DEBUG) |
| 1948 | { |
| 1949 | SanityManager.ASSERT(endPosition == LOG_FILE_HEADER_SIZE, |
| 1950 | "empty log file has wrong size"); |
| 1951 | } |
| 1952 | |
| 1953 | } |
| 1954 | else // something went wrong, delete the half baked file |
| 1955 | { |
| 1956 | newLog.close(); |
| 1957 | newLog = null; |
| 1958 | |
| 1959 | if (privExists(newLogFile)) |
| 1960 | privDelete(newLogFile); |
| 1961 | newLogFile = null; |
| 1962 | |
| 1963 | logErrMsg(MessageService.getTextMessage( |
| 1964 | MessageId.LOG_CANNOT_CREATE_NEW, |
| 1965 | newLogFile.getPath())); |
| 1966 | } |
| 1967 | |
| 1968 | } |
| 1969 | catch (IOException ioe) |
| 1970 | { |
| 1971 | |
| 1972 | inLogSwitch = false; |
| 1973 | // switching log file is an optional operation and there is no direct user |
| 1974 | // control. Just sends a warning message to whomever, if any, |
| 1975 | // system adminstrator there may be |
| 1976 | |
| 1977 | logErrMsg(MessageService.getTextMessage( |
| 1978 | MessageId.LOG_CANNOT_CREATE_NEW_DUETO, |
| 1979 | newLogFile.getPath(), |
| 1980 | ioe.toString())); |
| 1981 | |
| 1982 | try |
| 1983 | { |
| 1984 | if (newLog != null) |
| 1985 | { |
| 1986 | newLog.close(); |
| 1987 | newLog = null; |
| 1988 | } |
| 1989 | } |
| 1990 | catch (IOException ioe2) {} |
| 1991 | |
| 1992 | if (newLogFile != null && privExists(newLogFile)) |
| 1993 | { |
| 1994 | privDelete(newLogFile); |
| 1995 | newLogFile = null; |
| 1996 | } |
| 1997 | |
| 1998 | if (switchedOver) // error occur after old log file has been closed! |
| 1999 | { |
| 2000 | logOut = null; // limit any damage |
| 2001 | throw markCorrupt( |
| 2002 | StandardException.newException( |
| 2003 | SQLState.LOG_IO_ERROR, ioe)); |
| 2004 | } |
| 2005 | } |
| 2006 | |
| 2007 | inLogSwitch = false; |
| 2008 | } |
| 2009 | // unfreezes the log |
| 2010 | } |
| 2011 | |
| 2012 | /** |
| 2013 | Flush all unwritten log record up to the log instance indicated to disk |
| 2014 | without syncing. |
| 2015 | |
| 2016 | <P>MT - not needed, wrapper method |
| 2017 | |
| 2018 | @param wherePosition flush log up to here |
| 2019 | |
| 2020 | @exception IOException Failed to flush to the log |
| 2021 | */ |
| 2022 | private void flushBuffer(long fileNumber, long wherePosition) |
| 2023 | throws IOException, StandardException |
| 2024 | { |
| 2025 | synchronized (this) { |
| 2026 | if (fileNumber < logFileNumber) // history |
| 2027 | return; |
| 2028 | |
| 2029 | // A log instant indicates the start of a log record |
| 2030 | // but not how long it is. Thus the amount of data in |
| 2031 | // the logOut buffer is irrelevant. We can only |
| 2032 | // not flush the buffer if the real synced flush |
| 2033 | // included this required log instant. This is because |
| 2034 | // we never flush & sync partial log records. |
| 2035 | |
| 2036 | if (wherePosition < lastFlush) // already flushed |
| 2037 | return; |
| 2038 | |
| 2039 | // We don't update lastFlush here because lastFlush |
| 2040 | // is the last position in the log file that has been |
| 2041 | // flushed *and* synced to disk. Here we only flush. |
| 2042 | // ie. lastFlush should be renamed lastSync. |
| 2043 | // |
| 2044 | // We could have another variable indicating to which |
| 2045 | // point the log has been flushed which this routine |
| 2046 | // could take advantage of. This would only help rollbacks though. |
| 2047 | |
| 2048 | logOut.flushLogAccessFile(); |
| 2049 | } |
| 2050 | } |
| 2051 | /** Get rid of old and unnecessary log files |
| 2052 | |
| 2053 | <P> MT- only one truncate log is allowed to be taking place at any |
| 2054 | given time. Synchronized on this. |
| 2055 | |
| 2056 | */ |
| 2057 | private void truncateLog(CheckpointOperation checkpoint) |
| 2058 | { |
| 2059 | long oldFirstLog; |
| 2060 | long firstLogNeeded; |
| 2061 | |
| 2062 | if (keepAllLogs) |
| 2063 | return; |
| 2064 | if ((firstLogNeeded = getFirstLogNeeded(checkpoint))==-1) |
| 2065 | return; |
| 2066 | |
| 2067 | // when backup is in progress, log files that are yet to |
| 2068 | // be copied to the backup should not be deleted, even |
| 2069 | // if they are not required for crash recovery. |
| 2070 | if(backupInProgress) { |
| 2071 | long logFileNeededForBackup = logFileToBackup; |
| 2072 | // check if the log file number is yet to be copied |
| 2073 | // to the backup is less than the log file required |
| 2074 | // for crash recovery, if it is then make the first |
| 2075 | // log file that should not be deleted is the log file |
| 2076 | // that is yet to be copied to the backup. |
| 2077 | if (logFileNeededForBackup < firstLogNeeded) |
| 2078 | firstLogNeeded = logFileNeededForBackup; |
| 2079 | } |
| 2080 | |
| 2081 | oldFirstLog = firstLogFileNumber; |
| 2082 | firstLogFileNumber = firstLogNeeded; |
| 2083 | |
| 2084 | while(oldFirstLog < firstLogNeeded) |
| 2085 | { |
| 2086 | StorageFile uselessLogFile = null; |
| 2087 | try |
| 2088 | { |
| 2089 | uselessLogFile = getLogFileName(oldFirstLog); |
| 2090 | if (privDelete(uselessLogFile)) |
| 2091 | { |
| 2092 | if (SanityManager.DEBUG) |
| 2093 | { |
| 2094 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2095 | SanityManager.DEBUG(DBG_FLAG, "truncating useless log file " + uselessLogFile.getPath()); |
| 2096 | } |
| 2097 | } |
| 2098 | else |
| 2099 | { |
| 2100 | if (SanityManager.DEBUG) |
| 2101 | { |
| 2102 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2103 | SanityManager.DEBUG(DBG_FLAG, "Fail to truncate useless log file " + uselessLogFile.getPath()); |
| 2104 | } |
| 2105 | } |
| 2106 | } |
| 2107 | catch (StandardException se) |
| 2108 | { |
| 2109 | if (SanityManager.DEBUG) |
| 2110 | SanityManager.THROWASSERT("error opening log segment while deleting " |
| 2111 | + uselessLogFile.getPath(), se); |
| 2112 | |
| 2113 | // if insane, just leave it be |
| 2114 | } |
| 2115 | |
| 2116 | oldFirstLog++; |
| 2117 | } |
| 2118 | } |
| 2119 | |
| 2120 | |
| 2121 | |
| 2122 | private long getFirstLogNeeded(CheckpointOperation checkpoint){ |
| 2123 | |
| 2124 | long firstLogNeeded; |
| 2125 | |
| 2126 | // one truncation at a time |
| 2127 | synchronized (this) |
| 2128 | { |
| 2129 | firstLogNeeded = (checkpoint != null ? LogCounter.getLogFileNumber(checkpoint.undoLWM()) : -1); |
| 2130 | |
| 2131 | if (SanityManager.DEBUG) |
| 2132 | { |
| 2133 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2134 | SanityManager.DEBUG(DBG_FLAG, "truncatLog: undoLWM firstlog needed " + firstLogNeeded); |
| 2135 | } |
| 2136 | |
| 2137 | if (SanityManager.DEBUG) |
| 2138 | { |
| 2139 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2140 | { |
| 2141 | SanityManager.DEBUG(DBG_FLAG, "truncatLog: checkpoint truncationLWM firstlog needed " + firstLogNeeded); |
| 2142 | SanityManager.DEBUG(DBG_FLAG, "truncatLog: firstLogFileNumber = " + firstLogFileNumber); |
| 2143 | } |
| 2144 | } |
| 2145 | } |
| 2146 | return firstLogNeeded; |
| 2147 | } |
| 2148 | |
| 2149 | |
| 2150 | /** |
| 2151 | Carefully write out this value to the control file. |
| 2152 | We do safe write of this data by writing the data |
| 2153 | into two files every time we write the control data. |
| 2154 | we write checksum at the end of the file, so if by |
| 2155 | chance system crashes while writing into the file, |
| 2156 | using the checksum we find that the control file |
| 2157 | is hosed then we use the mirror file, which will have |
| 2158 | the condrol data written at last check point. |
| 2159 | |
| 2160 | see comment at beginning of file for log control file format. |
| 2161 | |
| 2162 | <P> MT- synchronized by caller |
| 2163 | */ |
| 2164 | boolean writeControlFile(StorageFile logControlFileName, long value) |
| 2165 | throws IOException, StandardException |
| 2166 | { |
| 2167 | StorageRandomAccessFile logControlFile = null; |
| 2168 | |
| 2169 | ByteArrayOutputStream baos = new ByteArrayOutputStream(64); |
| 2170 | DataOutputStream daos = new DataOutputStream(baos); |
| 2171 | |
| 2172 | daos.writeInt(fid); |
| 2173 | |
| 2174 | // so that when this db is booted by 1.1x and 1.2x JBMS, a IOException |
| 2175 | // stack trace rather than some error message that tells |
| 2176 | // the user to delete the database will show up. |
| 2177 | daos.writeInt(OBSOLETE_LOG_VERSION_NUMBER); |
| 2178 | daos.writeLong(value); |
| 2179 | |
| 2180 | if (onDiskMajorVersion == 0) { |
| 2181 | onDiskMajorVersion = jbmsVersion.getMajorVersion(); |
| 2182 | onDiskMinorVersion = jbmsVersion.getMinorVersion(); |
| 2183 | onDiskBeta = jbmsVersion.isBeta(); |
| 2184 | } |
| 2185 | |
| 2186 | // previous to 1.3, that's all we wrote. |
| 2187 | // from 1.3 and onward, also write out the JBMSVersion |
| 2188 | daos.writeInt(onDiskMajorVersion); |
| 2189 | daos.writeInt(onDiskMinorVersion); |
| 2190 | |
| 2191 | // For 2.0 beta we added the build number and the isBeta indication. |
| 2192 | // (5 bytes from our first spare long) |
| 2193 | daos.writeInt(jbmsVersion.getBuildNumberAsInt()); |
| 2194 | |
| 2195 | byte flags = 0; |
| 2196 | if (onDiskBeta) |
| 2197 | flags |= IS_BETA_FLAG; |
| 2198 | |
| 2199 | // When database is booted with derby.system.durability=test, |
| 2200 | // this mode does not guarantee that |
| 2201 | // - database will recover |
| 2202 | // - committed transactions will not be lost |
| 2203 | // - database will be in a consistent state |
| 2204 | // Hence necessary to keep track of this state so we don't |
| 2205 | // waste time resolving issues in such cases. |
| 2206 | // wasDBInDurabilityTestModeNoSync has information if database was |
| 2207 | // previously booted at any time in this mode |
| 2208 | if (logNotSynced || wasDBInDurabilityTestModeNoSync) |
| 2209 | flags |= IS_DURABILITY_TESTMODE_NO_SYNC_FLAG; |
| 2210 | daos.writeByte(flags); |
| 2211 | |
| 2212 | // |
| 2213 | // write some spare bytes after 2.0 we have 3 + 2(8) spare bytes. |
| 2214 | long spare = 0; |
| 2215 | |
| 2216 | daos.writeByte(0); |
| 2217 | daos.writeByte(0); |
| 2218 | daos.writeByte(0); |
| 2219 | daos.writeLong(spare); |
| 2220 | daos.flush(); |
| 2221 | // write the checksum for the control data written |
| 2222 | checksum.reset(); |
| 2223 | checksum.update(baos.toByteArray(), 0, baos.size()); |
| 2224 | daos.writeLong(checksum.getValue()); |
| 2225 | daos.flush(); |
| 2226 | |
| 2227 | try |
| 2228 | { |
| 2229 | checkCorrupt(); |
| 2230 | |
| 2231 | try |
| 2232 | { |
| 2233 | logControlFile = privRandomAccessFile(logControlFileName, "rw"); |
| 2234 | } |
| 2235 | catch (IOException ioe) |
| 2236 | { |
| 2237 | logControlFile = null; |
| 2238 | return false; |
| 2239 | } |
| 2240 | |
| 2241 | if (!privCanWrite(logControlFileName)) |
| 2242 | return false; |
| 2243 | |
| 2244 | if (SanityManager.DEBUG) |
| 2245 | { |
| 2246 | if (SanityManager.DEBUG_ON(TEST_LOG_FULL)) |
| 2247 | testLogFull(); |
| 2248 | } |
| 2249 | |
| 2250 | logControlFile.seek(0); |
| 2251 | logControlFile.write(baos.toByteArray()); |
| 2252 | syncFile(logControlFile); |
| 2253 | logControlFile.close(); |
| 2254 | |
| 2255 | // write the same data to mirror control file |
| 2256 | try |
| 2257 | { |
| 2258 | logControlFile = |
| 2259 | privRandomAccessFile(getMirrorControlFileName(), "rw"); |
| 2260 | } |
| 2261 | catch (IOException ioe) |
| 2262 | { |
| 2263 | logControlFile = null; |
| 2264 | return false; |
| 2265 | } |
| 2266 | |
| 2267 | logControlFile.seek(0); |
| 2268 | logControlFile.write(baos.toByteArray()); |
| 2269 | syncFile(logControlFile); |
| 2270 | |
| 2271 | } |
| 2272 | finally |
| 2273 | { |
| 2274 | if (logControlFile != null) |
| 2275 | logControlFile.close(); |
| 2276 | } |
| 2277 | |
| 2278 | return true; |
| 2279 | |
| 2280 | } |
| 2281 | |
| 2282 | /* |
| 2283 | Carefully read the content of the control file. |
| 2284 | |
| 2285 | <P> MT- read only |
| 2286 | */ |
| 2287 | private long readControlFile(StorageFile logControlFileName, Properties startParams) |
| 2288 | throws IOException, StandardException |
| 2289 | { |
| 2290 | StorageRandomAccessFile logControlFile = null; |
| 2291 | ByteArrayInputStream bais = null; |
| 2292 | DataInputStream dais = null; |
| 2293 | logControlFile = privRandomAccessFile(logControlFileName, "r"); |
| 2294 | boolean upgradeNeeded = false; |
| 2295 | long value = LogCounter.INVALID_LOG_INSTANT; |
| 2296 | long onDiskChecksum = 0; |
| 2297 | long controlFilelength = logControlFile.length(); |
| 2298 | byte barray[] = null; |
| 2299 | |
| 2300 | try |
| 2301 | { |
| 2302 | // The length of the file is less than the minimum in any version |
| 2303 | // It is possibly hosed , no point in reading data from this file |
| 2304 | // skip reading checksum control file is before 1.5 |
| 2305 | if (controlFilelength < 16) |
| 2306 | onDiskChecksum = -1; |
| 2307 | else if (controlFilelength == 16) |
| 2308 | { |
| 2309 | barray = new byte[16]; |
| 2310 | logControlFile.readFully(barray); |
| 2311 | }else if (controlFilelength > 16) |
| 2312 | { |
| 2313 | barray = new byte[(int) logControlFile.length() - 8]; |
| 2314 | logControlFile.readFully(barray); |
| 2315 | onDiskChecksum = logControlFile.readLong(); |
| 2316 | if (onDiskChecksum !=0 ) |
| 2317 | { |
| 2318 | checksum.reset(); |
| 2319 | checksum.update(barray, 0, barray.length); |
| 2320 | } |
| 2321 | } |
| 2322 | |
| 2323 | if ( onDiskChecksum == checksum.getValue() || onDiskChecksum ==0) |
| 2324 | { |
| 2325 | |
| 2326 | bais = new ByteArrayInputStream(barray); |
| 2327 | dais = new DataInputStream(bais); |
| 2328 | |
| 2329 | if (dais.readInt() != fid) |
| 2330 | { |
| 2331 | throw StandardException.newException( |
| 2332 | SQLState.LOG_INCOMPATIBLE_FORMAT, dataDirectory); |
| 2333 | } |
| 2334 | |
| 2335 | int obsoleteVersion = dais.readInt(); |
| 2336 | value = dais.readLong(); |
| 2337 | |
| 2338 | if (SanityManager.DEBUG) |
| 2339 | { |
| 2340 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2341 | SanityManager.DEBUG(LogToFile.DBG_FLAG, |
| 2342 | "log control file ckp instance = " + |
| 2343 | LogCounter.toDebugString(value)); |
| 2344 | } |
| 2345 | |
| 2346 | |
| 2347 | // from version 1.5 onward, we added an int for storing JBMS |
| 2348 | // version and an int for storing checkpoint interval |
| 2349 | // and log switch interval |
| 2350 | onDiskMajorVersion = dais.readInt(); |
| 2351 | onDiskMinorVersion = dais.readInt(); |
| 2352 | int dbBuildNumber = dais.readInt(); |
| 2353 | int flags = dais.readByte(); |
| 2354 | |
| 2355 | // check if the database was booted previously at any time with |
| 2356 | // derby.system.durability=test mode |
| 2357 | // If yes, then on a boot error we report that this setting is |
| 2358 | // probably the cause for the error and also log a warning |
| 2359 | // in the derby.log that this mode was set previously |
| 2360 | wasDBInDurabilityTestModeNoSync = |
| 2361 | (flags & IS_DURABILITY_TESTMODE_NO_SYNC_FLAG) != 0; |
| 2362 | |
| 2363 | if (SanityManager.DEBUG) { |
| 2364 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2365 | SanityManager.DEBUG(LogToFile.DBG_FLAG, |
| 2366 | "log control file, was derby.system.durability set to test = " + |
| 2367 | wasDBInDurabilityTestModeNoSync); |
| 2368 | } |
| 2369 | |
| 2370 | |
| 2371 | onDiskBeta = (flags & IS_BETA_FLAG) != 0; |
| 2372 | if (onDiskBeta) |
| 2373 | { |
| 2374 | // if is beta, can only be booted by exactly the same |
| 2375 | // version |
| 2376 | if (!jbmsVersion.isBeta() || |
| 2377 | onDiskMajorVersion != jbmsVersion.getMajorVersion() || |
| 2378 | onDiskMinorVersion != jbmsVersion.getMinorVersion()) |
| 2379 | { |
| 2380 | boolean forceBetaUpgrade = false; |
| 2381 | if (SanityManager.DEBUG) |
| 2382 | { |
| 2383 | // give ourselves an out for this beta check for debugging purposes |
| 2384 | if (SanityManager.DEBUG_ON("forceBetaUpgrade")) |
| 2385 | { |
| 2386 | Monitor.logMessage("WARNING !! : forcing beta upgrade."); |
| 2387 | forceBetaUpgrade =true; |
| 2388 | } |
| 2389 | } |
| 2390 | |
| 2391 | if (!forceBetaUpgrade) |
| 2392 | { |
| 2393 | throw StandardException.newException( |
| 2394 | SQLState.LOG_CANNOT_UPGRADE_BETA, |
| 2395 | dataDirectory, |
| 2396 | ProductVersionHolder.simpleVersionString(onDiskMajorVersion, onDiskMinorVersion, onDiskBeta)); |
| 2397 | } |
| 2398 | } |
| 2399 | } |
| 2400 | |
| 2401 | |
| 2402 | // JBMS_VERSION must be numbered in a way so that it is ever |
| 2403 | // increasing. We are backwards compatible but not forwards |
| 2404 | // compatible |
| 2405 | // |
| 2406 | if (onDiskMajorVersion > jbmsVersion.getMajorVersion() || |
| 2407 | (onDiskMajorVersion == jbmsVersion.getMajorVersion() && |
| 2408 | onDiskMinorVersion > jbmsVersion.getMinorVersion())) |
| 2409 | { |
| 2410 | // don't need to worry about point release, no format |
| 2411 | // upgrade is allowed. |
| 2412 | throw StandardException.newException( |
| 2413 | SQLState.LOG_INCOMPATIBLE_VERSION, |
| 2414 | dataDirectory, |
| 2415 | ProductVersionHolder.simpleVersionString(onDiskMajorVersion, onDiskMinorVersion, onDiskBeta)); |
| 2416 | } |
| 2417 | |
| 2418 | // Ensure that upgrade has been requested for a major or minor upgrade |
| 2419 | // maintaince (point) versions should not require an upgrade. |
| 2420 | if ((onDiskMajorVersion != jbmsVersion.getMajorVersion()) || |
| 2421 | (onDiskMinorVersion != jbmsVersion.getMinorVersion())) |
| 2422 | { |
| 2423 | upgradeNeeded = true; |
| 2424 | } |
| 2425 | // if checksum is zeros in version > 3.5 file is hosed |
| 2426 | // except incase of upgrade from versions <= 3.5 |
| 2427 | if (onDiskChecksum == 0 && |
| 2428 | (!(onDiskMajorVersion <= 3 && onDiskMinorVersion <=5) || |
| 2429 | onDiskMajorVersion == 0)) |
| 2430 | value = LogCounter.INVALID_LOG_INSTANT; |
| 2431 | } |
| 2432 | } |
| 2433 | finally |
| 2434 | { |
| 2435 | if (logControlFile != null) |
| 2436 | logControlFile.close(); |
| 2437 | if (bais != null) |
| 2438 | bais.close(); |
| 2439 | if (dais != null) |
| 2440 | dais.close(); |
| 2441 | } |
| 2442 | |
| 2443 | if (upgradeNeeded) |
| 2444 | { |
| 2445 | if (Monitor.isFullUpgrade(startParams, |
| 2446 | ProductVersionHolder.simpleVersionString(onDiskMajorVersion, onDiskMinorVersion, onDiskBeta))) { |
| 2447 | |
| 2448 | onDiskMajorVersion = jbmsVersion.getMajorVersion(); |
| 2449 | onDiskMinorVersion = jbmsVersion.getMinorVersion(); |
| 2450 | onDiskBeta = jbmsVersion.isBeta(); |
| 2451 | |
| 2452 | // Write out the new log control file with the new |
| 2453 | // version, the database has been upgraded |
| 2454 | |
| 2455 | if (!writeControlFile(logControlFileName, value)) |
| 2456 | { |
| 2457 | throw StandardException.newException( |
| 2458 | SQLState.LOG_CONTROL_FILE, logControlFileName); |
| 2459 | } |
| 2460 | } |
| 2461 | } |
| 2462 | |
| 2463 | return value; |
| 2464 | |
| 2465 | } |
| 2466 | |
| 2467 | |
| 2468 | |
| 2469 | /** |
| 2470 | * Create the directory where transaction log should go. |
| 2471 | * @exception StandardException Standard Error Policy |
| 2472 | */ |
| 2473 | private void createLogDirectory() throws StandardException |
| 2474 | { |
| 2475 | StorageFile logDir = |
| 2476 | logStorageFactory.newStorageFile(LogFactory.LOG_DIRECTORY_NAME); |
| 2477 | |
| 2478 | if (privExists(logDir)) { |
| 2479 | // make sure log directory is empty. |
| 2480 | String[] logfiles = privList(logDir); |
| 2481 | if (logfiles != null) { |
| 2482 | if(logfiles.length != 0) { |
| 2483 | throw StandardException.newException( |
| 2484 | SQLState.LOG_SEGMENT_EXIST, logDir.getPath()); |
| 2485 | } |
| 2486 | } |
| 2487 | |
| 2488 | }else { |
| 2489 | // create the log directory. |
| 2490 | if (!privMkdirs(logDir)) { |
| 2491 | throw StandardException.newException( |
| 2492 | SQLState.LOG_SEGMENT_NOT_EXIST, logDir.getPath()); |
| 2493 | } |
| 2494 | } |
| 2495 | } |
| 2496 | |
| 2497 | /* |
| 2498 | Return the directory the log should go. |
| 2499 | |
| 2500 | <P> MT- read only |
| 2501 | @exception StandardException Cloudscape Standard Error Policy |
| 2502 | */ |
| 2503 | public StorageFile getLogDirectory() throws StandardException |
| 2504 | { |
| 2505 | StorageFile logDir = null; |
| 2506 | |
| 2507 | logDir = logStorageFactory.newStorageFile( LogFactory.LOG_DIRECTORY_NAME); |
| 2508 | |
| 2509 | if (!privExists(logDir)) |
| 2510 | { |
| 2511 | throw StandardException.newException( |
| 2512 | SQLState.LOG_SEGMENT_NOT_EXIST, logDir.getPath()); |
| 2513 | } |
| 2514 | |
| 2515 | return logDir; |
| 2516 | } |
| 2517 | |
| 2518 | public String getCanonicalLogPath() |
| 2519 | { |
| 2520 | if (logDevice == null) |
| 2521 | return null; |
| 2522 | else |
| 2523 | { |
| 2524 | try |
| 2525 | { |
| 2526 | return logStorageFactory.getCanonicalName(); |
| 2527 | } |
| 2528 | catch (IOException ioe) |
| 2529 | { |
| 2530 | return null; |
| 2531 | // file not found |
| 2532 | } |
| 2533 | } |
| 2534 | } |
| 2535 | |
| 2536 | |
| 2537 | /** |
| 2538 | Return the control file name |
| 2539 | |
| 2540 | <P> MT- read only |
| 2541 | */ |
| 2542 | private StorageFile getControlFileName() throws StandardException |
| 2543 | { |
| 2544 | return logStorageFactory.newStorageFile( getLogDirectory(), "log.ctrl"); |
| 2545 | } |
| 2546 | |
| 2547 | /** |
| 2548 | Return the mirror control file name |
| 2549 | |
| 2550 | <P> MT- read only |
| 2551 | */ |
| 2552 | private StorageFile getMirrorControlFileName() throws StandardException |
| 2553 | { |
| 2554 | return logStorageFactory.newStorageFile( getLogDirectory(), "logmirror.ctrl"); |
| 2555 | } |
| 2556 | |
| 2557 | /** |
| 2558 | Given a log file number, return its file name |
| 2559 | |
| 2560 | <P> MT- read only |
| 2561 | */ |
| 2562 | private StorageFile getLogFileName(long filenumber) throws StandardException |
| 2563 | { |
| 2564 | return logStorageFactory.newStorageFile( getLogDirectory(), "log" + filenumber + ".dat"); |
| 2565 | } |
| 2566 | |
| 2567 | /* |
| 2568 | Find a checkpoint log record at the checkpointInstant |
| 2569 | |
| 2570 | <P> MT- read only |
| 2571 | */ |
| 2572 | private CheckpointOperation findCheckpoint(long checkpointInstant, FileLogger logger) |
| 2573 | throws IOException, StandardException, ClassNotFoundException |
| 2574 | { |
| 2575 | StreamLogScan scan = (StreamLogScan) |
| 2576 | openForwardsScan(checkpointInstant, (LogInstant)null); |
| 2577 | |
| 2578 | // estimated size of a checkpoint log record, which contains 3 longs |
| 2579 | // and assorted other log record overhead |
| 2580 | Loggable lop = logger.readLogRecord(scan, 100); |
| 2581 | |
| 2582 | scan.close(); |
| 2583 | |
| 2584 | if (lop instanceof CheckpointOperation) |
| 2585 | return (CheckpointOperation)lop; |
| 2586 | else |
| 2587 | return null; |
| 2588 | } |
| 2589 | |
| 2590 | |
| 2591 | /* |
| 2592 | * Functions to help the Logger open a log scan on the log. |
| 2593 | */ |
| 2594 | |
| 2595 | /** |
| 2596 | Scan backward from start position. |
| 2597 | |
| 2598 | <P> MT- read only |
| 2599 | |
| 2600 | @exception IOException cannot access the log |
| 2601 | @exception StandardException Standard Cloudscape error policy |
| 2602 | */ |
| 2603 | protected LogScan openBackwardsScan(long startAt, LogInstant stopAt) |
| 2604 | throws IOException, StandardException |
| 2605 | { |
| 2606 | checkCorrupt(); |
| 2607 | |
| 2608 | // backward from end of log |
| 2609 | if (startAt == LogCounter.INVALID_LOG_INSTANT) |
| 2610 | return openBackwardsScan(stopAt); |
| 2611 | |
| 2612 | |
| 2613 | // ensure any buffered data is written to the actual file |
| 2614 | flushBuffer(LogCounter.getLogFileNumber(startAt), |
| 2615 | LogCounter.getLogFilePosition(startAt)); |
| 2616 | |
| 2617 | return new Scan(this, startAt, stopAt, Scan.BACKWARD); |
| 2618 | } |
| 2619 | |
| 2620 | /** |
| 2621 | Scan backward from end of log. |
| 2622 | <P> MT- read only |
| 2623 | |
| 2624 | @exception IOException cannot access the log |
| 2625 | @exception StandardException Standard Cloudscape error policy |
| 2626 | */ |
| 2627 | protected LogScan openBackwardsScan(LogInstant stopAt) |
| 2628 | throws IOException, StandardException |
| 2629 | { |
| 2630 | checkCorrupt(); |
| 2631 | |
| 2632 | // current instant log instant of the next log record to be |
| 2633 | // written out, which is at the end of the log |
| 2634 | // ensure any buffered data is written to the actual file |
| 2635 | long startAt; |
| 2636 | synchronized (this) |
| 2637 | { |
| 2638 | // flush the whole buffer to ensure the complete |
| 2639 | // end of log is in the file. |
| 2640 | logOut.flushLogAccessFile(); |
| 2641 | startAt = currentInstant(); |
| 2642 | } |
| 2643 | |
| 2644 | return new Scan(this, startAt, stopAt, Scan.BACKWARD_FROM_LOG_END); |
| 2645 | } |
| 2646 | |
| 2647 | /** |
| 2648 | @see LogFactory#openFlushedScan |
| 2649 | @exception StandardException Ooops. |
| 2650 | */ |
| 2651 | public ScanHandle openFlushedScan(DatabaseInstant start,int groupsIWant) |
| 2652 | throws StandardException |
| 2653 | { |
| 2654 | return new FlushedScanHandle(this,start,groupsIWant); |
| 2655 | } |
| 2656 | |
| 2657 | |
| 2658 | |
| 2659 | /** |
| 2660 | Scan Forward from start position. |
| 2661 | |
| 2662 | <P> MT- read only |
| 2663 | |
| 2664 | @param startAt - if startAt == INVALID_LOG_INSTANT, |
| 2665 | start from the beginning of the log. Otherwise, start scan from startAt. |
| 2666 | @param stopAt - if not null, stop at this log instant (inclusive). |
| 2667 | Otherwise, stop at the end of the log |
| 2668 | |
| 2669 | @exception IOException cannot access the log |
| 2670 | @exception StandardException Standard Cloudscape error policy |
| 2671 | */ |
| 2672 | protected LogScan openForwardsScan(long startAt, LogInstant stopAt) |
| 2673 | throws IOException, StandardException |
| 2674 | { |
| 2675 | checkCorrupt(); |
| 2676 | |
| 2677 | if (startAt == LogCounter.INVALID_LOG_INSTANT) |
| 2678 | { |
| 2679 | startAt = firstLogInstant(); |
| 2680 | } |
| 2681 | |
| 2682 | // ensure any buffered data is written to the actual file |
| 2683 | if (stopAt != null) { |
| 2684 | LogCounter stopCounter = (LogCounter) stopAt; |
| 2685 | flushBuffer(stopCounter.getLogFileNumber(), |
| 2686 | stopCounter.getLogFilePosition()); |
| 2687 | } else { |
| 2688 | synchronized (this) { |
| 2689 | if (logOut != null) |
| 2690 | // flush to the end of the log |
| 2691 | logOut.flushLogAccessFile(); |
| 2692 | } |
| 2693 | } |
| 2694 | |
| 2695 | return new Scan(this, startAt, stopAt, Scan.FORWARD); |
| 2696 | } |
| 2697 | |
| 2698 | /* |
| 2699 | * Methods to help a log scan switch from one log file to the next |
| 2700 | */ |
| 2701 | |
| 2702 | /** |
| 2703 | Open a log file and position the file at the beginning. |
| 2704 | Used by scan to switch to the next log file |
| 2705 | |
| 2706 | <P> MT- read only |
| 2707 | |
| 2708 | @exception StandardException Standard Cloudscape error policy |
| 2709 | @exception IOException cannot access the log at the new position. |
| 2710 | */ |
| 2711 | protected StorageRandomAccessFile getLogFileAtBeginning(long filenumber) |
| 2712 | throws IOException, StandardException |
| 2713 | { |
| 2714 | long instant = LogCounter.makeLogInstantAsLong(filenumber, |
| 2715 | LOG_FILE_HEADER_SIZE); |
| 2716 | return getLogFileAtPosition(instant); |
| 2717 | } |
| 2718 | |
| 2719 | |
| 2720 | /** |
| 2721 | Get a read-only handle to the log file positioned at the stated position |
| 2722 | |
| 2723 | <P> MT- read only |
| 2724 | |
| 2725 | @return null if file does not exist or of the wrong format |
| 2726 | @exception IOException cannot access the log at the new position. |
| 2727 | @exception StandardException Standard Cloudscape error policy |
| 2728 | */ |
| 2729 | protected StorageRandomAccessFile getLogFileAtPosition(long logInstant) |
| 2730 | throws IOException, StandardException |
| 2731 | { |
| 2732 | checkCorrupt(); |
| 2733 | |
| 2734 | long filenum = LogCounter.getLogFileNumber(logInstant); |
| 2735 | long filepos = LogCounter.getLogFilePosition(logInstant); |
| 2736 | |
| 2737 | StorageFile fileName = getLogFileName(filenum); |
| 2738 | if (!privExists(fileName)) |
| 2739 | { |
| 2740 | if (SanityManager.DEBUG) |
| 2741 | { |
| 2742 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2743 | SanityManager.DEBUG(LogToFile.DBG_FLAG, fileName.getPath() + " does not exist"); |
| 2744 | } |
| 2745 | |
| 2746 | return null; |
| 2747 | } |
| 2748 | |
| 2749 | |
| 2750 | StorageRandomAccessFile log = null; |
| 2751 | |
| 2752 | try |
| 2753 | { |
| 2754 | log = privRandomAccessFile(fileName, "r"); |
| 2755 | |
| 2756 | // verify that the log file is of the right format |
| 2757 | if (!verifyLogFormat(log, filenum)) |
| 2758 | { |
| 2759 | if (SanityManager.DEBUG) |
| 2760 | { |
| 2761 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 2762 | SanityManager.DEBUG(LogToFile.DBG_FLAG, fileName.getPath() + " format mismatch"); |
| 2763 | } |
| 2764 | |
| 2765 | log.close(); |
| 2766 | log = null; |
| 2767 | } |
| 2768 | else |
| 2769 | { |
| 2770 | log.seek(filepos); |
| 2771 | } |
| 2772 | } |
| 2773 | catch (IOException ioe) |
| 2774 | { |
| 2775 | try |
| 2776 | { |
| 2777 | if (log != null) |
| 2778 | { |
| 2779 | log.close(); |
| 2780 | log = null; |
| 2781 | } |
| 2782 | |
| 2783 | if (SanityManager.DEBUG) |
| 2784 | { |
| 2785 | SanityManager.THROWASSERT("cannot get to position " + filepos + |
| 2786 | " for log file " + fileName.getPath(), ioe); |
| 2787 | } |
| 2788 | } |
| 2789 | catch (IOException ioe2) |
| 2790 | {} |
| 2791 | throw ioe; |
| 2792 | } |
| 2793 | |
| 2794 | return log; |
| 2795 | |
| 2796 | } |
| 2797 | |
| 2798 | /* |
| 2799 | ** Methods of ModuleControl |
| 2800 | */ |
| 2801 | |
| 2802 | public boolean canSupport(Properties startParams) |
| 2803 | { |
| 2804 | String runtimeLogAttributes = startParams.getProperty(LogFactory.RUNTIME_ATTRIBUTES); |
| 2805 | if (runtimeLogAttributes != null) { |
| 2806 | if (runtimeLogAttributes.equals(LogFactory.RT_READONLY)) |
| 2807 | return false; |
| 2808 | } |
| 2809 | |
| 2810 | return true; |
| 2811 | } |
| 2812 | |
| 2813 | |
| 2814 | |
| 2815 | |
| 2816 | /** |
| 2817 | Boot up the log factory. |
| 2818 | <P> MT- caller provide synchronization |
| 2819 | |
| 2820 | @exception StandardException log factory cannot start up |
| 2821 | */ |
| 2822 | public void boot(boolean create, Properties startParams) throws StandardException |
| 2823 | { |
| 2824 | dataDirectory = startParams.getProperty(PersistentService.ROOT); |
| 2825 | |
| 2826 | logDevice = startParams.getProperty(Attribute.LOG_DEVICE); |
| 2827 | if( logDevice != null) |
| 2828 | { |
| 2829 | // in case the user specifies logDevice in URL form |
| 2830 | String logDeviceURL = null; |
| 2831 | try { |
| 2832 | URL url = new URL(logDevice); |
| 2833 | logDeviceURL = url.getFile(); |
| 2834 | } catch (MalformedURLException ex) {} |
| 2835 | if (logDeviceURL != null) |
| 2836 | logDevice = logDeviceURL; |
| 2837 | } |
| 2838 | |
| 2839 | |
| 2840 | if(create) { |
| 2841 | getLogStorageFactory(); |
| 2842 | createLogDirectory(); |
| 2843 | |
| 2844 | } else { |
| 2845 | // check if the database is being restored from the backup, |
| 2846 | // if it is then restore the logs. |
| 2847 | if (!restoreLogs(startParams)) { |
| 2848 | // set the log storage factory. |
| 2849 | getLogStorageFactory(); |
| 2850 | if (logDevice != null) |
| 2851 | { |
| 2852 | // Make sure we find the log, do not assume |
| 2853 | // it is OK that the log is not there because |
| 2854 | // it could be a user typo(like when users edit |
| 2855 | // service.properties to change the log device |
| 2856 | // while restoring from backups using OS copy. |
| 2857 | StorageFile checklogDir = |
| 2858 | logStorageFactory.newStorageFile( |
| 2859 | LogFactory.LOG_DIRECTORY_NAME); |
| 2860 | if (!privExists(checklogDir)) |
| 2861 | { |
| 2862 | throw |
| 2863 | StandardException.newException( |
| 2864 | SQLState.LOG_FILE_NOT_FOUND, checklogDir.getPath()); |
| 2865 | |
| 2866 | } |
| 2867 | } |
| 2868 | } |
| 2869 | } |
| 2870 | |
| 2871 | //if user does not set the right value for the log buffer size, |
| 2872 | //default value is used instead. |
| 2873 | logBufferSize = PropertyUtil.getSystemInt(org.apache.derby.iapi.reference.Property.LOG_BUFFER_SIZE, |
| 2874 | LOG_BUFFER_SIZE_MIN, |
| 2875 | LOG_BUFFER_SIZE_MAX, |
| 2876 | DEFAULT_LOG_BUFFER_SIZE); |
| 2877 | jbmsVersion = Monitor.getMonitor().getEngineVersion(); |
| 2878 | |
| 2879 | |
| 2880 | String logArchiveMode = |
| 2881 | startParams.getProperty(Property.LOG_ARCHIVE_MODE); |
| 2882 | logArchived = Boolean.valueOf(logArchiveMode).booleanValue(); |
| 2883 | |
| 2884 | //get log factorty properties if any set in derby.properties |
| 2885 | getLogFactoryProperties(null); |
| 2886 | |
| 2887 | /* check if the storage factory supports write sync(rws). If so, use it unless |
| 2888 | * derby.storage.fileSyncTransactionLog property is set true by user. |
| 2889 | */ |
| 2890 | |
| 2891 | if (logStorageFactory.supportsRws()) |
| 2892 | { |
| 2893 | //write sync can be used in the jvm that database is running on. |
| 2894 | //disable write sync if derby.storage.fileSyncTransactionLog is true |
| 2895 | isWriteSynced = |
| 2896 | !(PropertyUtil.getSystemBoolean(Property.FILESYNC_TRANSACTION_LOG)); |
| 2897 | } |
| 2898 | else |
| 2899 | { |
| 2900 | isWriteSynced = false; |
| 2901 | } |
| 2902 | |
| 2903 | |
| 2904 | // If derby.system.durability=test is set,then set flag to |
| 2905 | // disable sync of log records at commit and log file before |
| 2906 | // data page makes it to disk |
| 2907 | if (Property.DURABILITY_TESTMODE_NO_SYNC.equalsIgnoreCase( |
| 2908 | PropertyUtil.getSystemProperty(Property.DURABILITY_PROPERTY))) |
| 2909 | { |
| 2910 | // disable syncing of log. |
| 2911 | logNotSynced = true; |
| 2912 | //if log not being synced;files shouldn't be open in write sync mode |
| 2913 | isWriteSynced = false; |
| 2914 | } |
| 2915 | else if (Performance.MEASURE) |
| 2916 | { |
| 2917 | // development build only feature, must by hand set the |
| 2918 | // Performance.MEASURE variable and rebuild. Useful during |
| 2919 | // development to compare/contrast effect of syncing, release |
| 2920 | // users can use the above relaxed durability option to disable |
| 2921 | // all syncing. |
| 2922 | |
| 2923 | logNotSynced = |
| 2924 | PropertyUtil.getSystemBoolean( |
| 2925 | Property.STORAGE_LOG_NOT_SYNCED); |
| 2926 | |
| 2927 | if (logNotSynced) |
| 2928 | { |
| 2929 | isWriteSynced = false; |
| 2930 | Monitor.logMessage("Performance.logNotSynced = true"); |
| 2931 | } |
| 2932 | } |
| 2933 | |
| 2934 | // try to access the log |
| 2935 | // if it doesn't exist, create it. |
| 2936 | // if it does exist, run recovery |
| 2937 | |
| 2938 | boolean createNewLog = create; |
| 2939 | |
| 2940 | if (SanityManager.DEBUG) |
| 2941 | SanityManager.ASSERT(fid != -1, "invalid log format Id"); |
| 2942 | |
| 2943 | checkpointInstant = LogCounter.INVALID_LOG_INSTANT; |
| 2944 | try |
| 2945 | { |
| 2946 | StorageFile logControlFileName = getControlFileName(); |
| 2947 | |
| 2948 | StorageFile logFile; |
| 2949 | |
| 2950 | if (!createNewLog) |
| 2951 | { |
| 2952 | if (privExists(logControlFileName)) |
| 2953 | { |
| 2954 | checkpointInstant = |
| 2955 | readControlFile(logControlFileName, startParams); |
| 2956 | |
| 2957 | // in case system was running previously with |
| 2958 | // derby.system.durability=test then print a message |
| 2959 | // to the derby log |
| 2960 | if (wasDBInDurabilityTestModeNoSync) |
| 2961 | { |
| 2962 | // print message stating that the database was |
| 2963 | // previously atleast at one time running with |
| 2964 | // derby.system.durability=test mode |
| 2965 | Monitor.logMessage(MessageService.getTextMessage( |
| 2966 | MessageId.LOG_WAS_IN_DURABILITY_TESTMODE_NO_SYNC, |
| 2967 | Property.DURABILITY_PROPERTY, |
| 2968 | Property.DURABILITY_TESTMODE_NO_SYNC)); |
| 2969 | } |
| 2970 | |
| 2971 | if (checkpointInstant == LogCounter.INVALID_LOG_INSTANT && |
| 2972 | getMirrorControlFileName().exists()) |
| 2973 | { |
| 2974 | checkpointInstant = |
| 2975 | readControlFile( |
| 2976 | getMirrorControlFileName(), startParams); |
| 2977 | } |
| 2978 | |
| 2979 | } |
| 2980 | else if (logDevice != null) |
| 2981 | { |
| 2982 | // Do not throw this error if logDevice is null because |
| 2983 | // in a read only configuration, it is acceptable |
| 2984 | // to not have a log directory. But clearly, if the |
| 2985 | // logDevice property is set, then it should be there. |
| 2986 | throw StandardException.newException( |
| 2987 | SQLState.LOG_FILE_NOT_FOUND, |
| 2988 | logControlFileName.getPath()); |
| 2989 | } |
| 2990 | |
| 2991 | if (checkpointInstant != LogCounter.INVALID_LOG_INSTANT) |
| 2992 | logFileNumber = LogCounter.getLogFileNumber(checkpointInstant); |
| 2993 | else |
| 2994 | logFileNumber = 1; |
| 2995 | |
| 2996 | logFile = getLogFileName(logFileNumber); |
| 2997 | |
| 2998 | // if log file is not there or if it is of the wrong format, create a |
| 2999 | // brand new log file and do not attempt to recover the database |
| 3000 | |
| 3001 | if (!privExists(logFile)) |
| 3002 | { |
| 3003 | if (logDevice != null) |
| 3004 | { |
| 3005 | throw StandardException.newException( |
| 3006 | SQLState.LOG_FILE_NOT_FOUND, |
| 3007 | logControlFileName.getPath()); |
| 3008 | } |
| 3009 | |
| 3010 | logErrMsg(MessageService.getTextMessage( |
| 3011 | MessageId.LOG_MAYBE_INCONSISTENT, |
| 3012 | logFile.getPath())); |
| 3013 | |
| 3014 | createNewLog = true; |
| 3015 | } |
| 3016 | else if (!verifyLogFormat(logFile, logFileNumber)) |
| 3017 | { |
| 3018 | Monitor.logTextMessage(MessageId.LOG_DELETE_INCOMPATIBLE_FILE, logFile); |
| 3019 | |
| 3020 | // blow away the log file if possible |
| 3021 | if (!privDelete(logFile) && logFileNumber == 1) |
| 3022 | { |
| 3023 | logErrMsgForDurabilityTestModeNoSync(); |
| 3024 | throw StandardException.newException( |
| 3025 | SQLState.LOG_INCOMPATIBLE_FORMAT, dataDirectory); |
| 3026 | } |
| 3027 | |
| 3028 | // If logFileNumber > 1, we are not going to write that |
| 3029 | // file just yet. Just leave it be and carry on. Maybe |
| 3030 | // when we get there it can be deleted. |
| 3031 | |
| 3032 | createNewLog = true; |
| 3033 | } |
| 3034 | } |
| 3035 | |
| 3036 | if (createNewLog) |
| 3037 | { |
| 3038 | // brand new log. Start from log file number 1. |
| 3039 | |
| 3040 | // create or overwrite the log control file with an invalid |
| 3041 | // checkpoint instant since there is no checkpoint yet |
| 3042 | if (writeControlFile(logControlFileName, |
| 3043 | LogCounter.INVALID_LOG_INSTANT)) |
| 3044 | { |
| 3045 | firstLogFileNumber = 1; |
| 3046 | logFileNumber = 1; |
| 3047 | if (SanityManager.DEBUG) |
| 3048 | { |
| 3049 | if (SanityManager.DEBUG_ON(TEST_MAX_LOGFILE_NUMBER)) |
| 3050 | { |
| 3051 | // set the value to be two less than max possible |
| 3052 | // log number, test case will perform some ops to |
| 3053 | // hit the max number case. |
| 3054 | firstLogFileNumber = |
| 3055 | LogCounter.MAX_LOGFILE_NUMBER -2; |
| 3056 | |
| 3057 | logFileNumber = LogCounter.MAX_LOGFILE_NUMBER -2; |
| 3058 | } |
| 3059 | } |
| 3060 | logFile = getLogFileName(logFileNumber); |
| 3061 | |
| 3062 | if (privExists(logFile)) |
| 3063 | { |
| 3064 | // this log file maybe there because the system may have |
| 3065 | // crashed right after a log switch but did not write |
| 3066 | // out any log record |
| 3067 | Monitor.logTextMessage( |
| 3068 | MessageId.LOG_DELETE_OLD_FILE, logFile); |
| 3069 | |
| 3070 | if (!privDelete(logFile)) |
| 3071 | { |
| 3072 | logErrMsgForDurabilityTestModeNoSync(); |
| 3073 | throw StandardException.newException( |
| 3074 | SQLState.LOG_INCOMPATIBLE_FORMAT, |
| 3075 | dataDirectory); |
| 3076 | } |
| 3077 | } |
| 3078 | |
| 3079 | // don't need to try to delete it, we know it isn't there |
| 3080 | firstLog = privRandomAccessFile(logFile, "rw"); |
| 3081 | |
| 3082 | if (!initLogFile(firstLog, logFileNumber, LogCounter.INVALID_LOG_INSTANT)) |
| 3083 | { |
| 3084 | throw StandardException.newException( |
| 3085 | SQLState.LOG_SEGMENT_NOT_EXIST, logFile.getPath()); |
| 3086 | } |
| 3087 | |
| 3088 | endPosition = firstLog.getFilePointer(); |
| 3089 | lastFlush = firstLog.getFilePointer(); |
| 3090 | |
| 3091 | //if write sync is true , prellocate the log file |
| 3092 | //and reopen the file in rws mode. |
| 3093 | if(isWriteSynced) |
| 3094 | { |
| 3095 | //extend the file by wring zeros to it |
| 3096 | preAllocateNewLogFile(firstLog); |
| 3097 | firstLog.close(); |
| 3098 | firstLog = openLogFileInWriteMode(logFile); |
| 3099 | //postion the log at the current log end postion |
| 3100 | firstLog.seek(endPosition); |
| 3101 | } |
| 3102 | |
| 3103 | if (SanityManager.DEBUG) |
| 3104 | { |
| 3105 | SanityManager.ASSERT( |
| 3106 | endPosition == LOG_FILE_HEADER_SIZE, |
| 3107 | "empty log file has wrong size"); |
| 3108 | } |
| 3109 | } |
| 3110 | else |
| 3111 | { |
| 3112 | // read only database |
| 3113 | ReadOnlyDB = true; |
| 3114 | logOut = null; |
| 3115 | firstLog = null; |
| 3116 | } |
| 3117 | |
| 3118 | recoveryNeeded = false; |
| 3119 | } |
| 3120 | else |
| 3121 | { |
| 3122 | // log file exist, need to run recovery |
| 3123 | recoveryNeeded = true; |
| 3124 | } |
| 3125 | |
| 3126 | } |
| 3127 | catch (IOException ioe) |
| 3128 | { |
| 3129 | throw Monitor.exceptionStartingModule(ioe); |
| 3130 | } |
| 3131 | |
| 3132 | // Number of the log file that can be created in Derby is increased from |
| 3133 | // 2^22 -1 to 2^31 -1 in version 10.1. But if the database is running on |
| 3134 | // engines 10.1 or above on a softupgrade from versions 10.0 or |
| 3135 | // before, the max log file number that can be created is |
| 3136 | // still limited to 2^22 -1, because users can revert back to older versions |
| 3137 | // which does not have logic to handle a log file number greater than |
| 3138 | // 2^22-1. |
| 3139 | |
| 3140 | // set max possible log file number to derby 10.0 limit, if the database is not |
| 3141 | // fully upgraded to or created in version 10.1 or above. |
| 3142 | if (!checkVersion(RawStoreFactory.DERBY_STORE_MAJOR_VERSION_10, |
| 3143 | RawStoreFactory.DERBY_STORE_MINOR_VERSION_1)) |
| 3144 | maxLogFileNumber = LogCounter.DERBY_10_0_MAX_LOGFILE_NUMBER; |
| 3145 | |
| 3146 | } // end of boot |
| 3147 | |
| 3148 | private void getLogStorageFactory() throws StandardException |
| 3149 | { |
| 3150 | if( logDevice == null) |
| 3151 | { |
| 3152 | DataFactory df = (DataFactory) Monitor.findServiceModule( this, DataFactory.MODULE); |
| 3153 | logStorageFactory = (WritableStorageFactory) df.getStorageFactory(); |
| 3154 | } |
| 3155 | else |
| 3156 | { |
| 3157 | try |
| 3158 | { |
| 3159 | PersistentService ps = Monitor.getMonitor().getServiceType(this); |
| 3160 | logStorageFactory = (WritableStorageFactory) ps.getStorageFactoryInstance( false, logDevice, null, null); |
| 3161 | } |
| 3162 | catch( IOException ioe) |
| 3163 | { |
| 3164 | if( SanityManager.DEBUG) |
| 3165 | SanityManager.NOTREACHED(); |
| 3166 | throw StandardException.newException( SQLState.LOG_FILE_NOT_FOUND, ioe, logDevice); |
| 3167 | } |
| 3168 | } |
| 3169 | } // end of getLogStorageFactory |
| 3170 | |
| 3171 | /** |
| 3172 | Stop the log factory |
| 3173 | <P> MT- caller provide synchronization |
| 3174 | (RESOLVE: this should be called AFTER dataFactory and transFactory are |
| 3175 | stopped) |
| 3176 | */ |
| 3177 | public void stop() { |
| 3178 | |
| 3179 | |
| 3180 | // stop our checkpoint |
| 3181 | if (checkpointDaemon != null) { |
| 3182 | checkpointDaemon.unsubscribe(myClientNumber); |
| 3183 | checkpointDaemon.stop(); |
| 3184 | } |
| 3185 | |
| 3186 | synchronized(this) |
| 3187 | { |
| 3188 | stopped = true; |
| 3189 | |
| 3190 | if (logOut != null) { |
| 3191 | try { |
| 3192 | logOut.flushLogAccessFile(); |
| 3193 | logOut.close(); |
| 3194 | } |
| 3195 | catch (IOException ioe) {} |
| 3196 | catch(StandardException se){} |
| 3197 | logOut = null; |
| 3198 | } |
| 3199 | } |
| 3200 | |
| 3201 | |
| 3202 | if (SanityManager.DEBUG && |
| 3203 | Performance.MEASURE && |
| 3204 | mon_LogSyncStatistics) |
| 3205 | { |
| 3206 | Monitor.logMessage("number of times someone waited = " + |
| 3207 | mon_numLogFlushWaits + |
| 3208 | "\nnumber of times flush is called = " + |
| 3209 | mon_flushCalls + |
| 3210 | "\nnumber of sync is called = " + |
| 3211 | mon_syncCalls + |
| 3212 | "\ntotal number of bytes written to log = " + |
| 3213 | LogAccessFile.mon_numBytesToLog + |
| 3214 | "\ntotal number of writes to log file = " + |
| 3215 | LogAccessFile.mon_numWritesToLog); |
| 3216 | } |
| 3217 | |
| 3218 | |
| 3219 | // delete obsolete log files,left around by earlier crashes |
| 3220 | if(corrupt == null && ! logArchived() && !keepAllLogs && !ReadOnlyDB) |
| 3221 | deleteObsoleteLogfiles(); |
| 3222 | |
| 3223 | if( logDevice != null) |
| 3224 | logStorageFactory.shutdown(); |
| 3225 | logStorageFactory = null; |
| 3226 | } |
| 3227 | |
| 3228 | |
| 3229 | |
| 3230 | /* delete the log files, that might have been left around if we crashed |
| 3231 | * immediately after the checkpoint before truncations of logs completed. |
| 3232 | * see bug no: 3519 , for more details. |
| 3233 | */ |
| 3234 | |
| 3235 | private void deleteObsoleteLogfiles(){ |
| 3236 | StorageFile logDir; |
| 3237 | //find the first log file number that is useful |
| 3238 | long firstLogNeeded = getFirstLogNeeded(currentCheckpoint); |
| 3239 | if (firstLogNeeded == -1) |
| 3240 | return; |
| 3241 | |
| 3242 | // when backup is in progress, log files that are yet to |
| 3243 | // be copied to the backup should not be deleted, even |
| 3244 | // if they are not required for crash recovery. |
| 3245 | if(backupInProgress) { |
| 3246 | long logFileNeededForBackup = logFileToBackup; |
| 3247 | // check if the log file number is yet to be copied |
| 3248 | // to the backup is less than the log file required |
| 3249 | // for crash recovery, if it is then make the first |
| 3250 | // log file that should not be deleted is the log file |
| 3251 | // that is yet to be copied to the backup. |
| 3252 | if (logFileNeededForBackup < firstLogNeeded) |
| 3253 | firstLogNeeded = logFileNeededForBackup; |
| 3254 | } |
| 3255 | |
| 3256 | try{ |
| 3257 | logDir = getLogDirectory(); |
| 3258 | }catch (StandardException se) |
| 3259 | { |
| 3260 | if (SanityManager.DEBUG) |
| 3261 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 3262 | SanityManager.DEBUG(DBG_FLAG, "error opening log segment dir"); |
| 3263 | return; |
| 3264 | } |
| 3265 | |
| 3266 | String[] logfiles = privList(logDir); |
| 3267 | if (logfiles != null) |
| 3268 | { |
| 3269 | StorageFile uselessLogFile = null; |
| 3270 | long fileNumber; |
| 3271 | for(int i=0 ; i < logfiles.length; i++) |
| 3272 | { |
| 3273 | // delete the log files that are not needed any more |
| 3274 | if(logfiles[i].startsWith("log") && logfiles[i].endsWith(".dat")) |
| 3275 | { |
| 3276 | fileNumber = Long.parseLong(logfiles[i].substring(3, (logfiles[i].length() -4))); |
| 3277 | if(fileNumber < firstLogNeeded ) |
| 3278 | { |
| 3279 | uselessLogFile = logStorageFactory.newStorageFile(logDir, logfiles[i]); |
| 3280 | if (privDelete(uselessLogFile)) |
| 3281 | { |
| 3282 | if (SanityManager.DEBUG) |
| 3283 | { |
| 3284 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 3285 | SanityManager.DEBUG(DBG_FLAG, "truncating obsolete log file " + uselessLogFile.getPath()); |
| 3286 | } |
| 3287 | } |
| 3288 | else |
| 3289 | { |
| 3290 | if (SanityManager.DEBUG) |
| 3291 | { |
| 3292 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 3293 | SanityManager.DEBUG(DBG_FLAG, "Fail to truncate obsolete log file " + uselessLogFile.getPath()); |
| 3294 | } |
| 3295 | } |
| 3296 | } |
| 3297 | } |
| 3298 | } |
| 3299 | } |
| 3300 | } |
| 3301 | |
| 3302 | /* |
| 3303 | * Serviceable methods |
| 3304 | */ |
| 3305 | |
| 3306 | public boolean serviceASAP() |
| 3307 | { |
| 3308 | return false; |
| 3309 | } |
| 3310 | |
| 3311 | // @return true, if this work needs to be done on a user thread immediately |
| 3312 | public boolean serviceImmediately() |
| 3313 | { |
| 3314 | return false; |
| 3315 | } |
| 3316 | |
| 3317 | |
| 3318 | public void getLogFactoryProperties(PersistentSet set) |
| 3319 | throws StandardException |
| 3320 | { |
| 3321 | String lsInterval; |
| 3322 | String cpInterval; |
| 3323 | if(set == null) |
| 3324 | { |
| 3325 | lsInterval=PropertyUtil.getSystemProperty(org.apache.derby.iapi.reference.Property.LOG_SWITCH_INTERVAL); |
| 3326 | cpInterval=PropertyUtil.getSystemProperty(org.apache.derby.iapi.reference.Property.CHECKPOINT_INTERVAL); |
| 3327 | }else |
| 3328 | { |
| 3329 | lsInterval = PropertyUtil.getServiceProperty(set, org.apache.derby.iapi.reference.Property.LOG_SWITCH_INTERVAL); |
| 3330 | cpInterval = PropertyUtil.getServiceProperty(set, org.apache.derby.iapi.reference.Property.CHECKPOINT_INTERVAL); |
| 3331 | } |
| 3332 | |
| 3333 | /* log switch interval */ |
| 3334 | if (lsInterval != null) |
| 3335 | { |
| 3336 | logSwitchInterval = Integer.parseInt(lsInterval); |
| 3337 | // make sure checkpoint and log switch interval are within range |
| 3338 | if (logSwitchInterval < LOG_SWITCH_INTERVAL_MIN) |
| 3339 | logSwitchInterval = LOG_SWITCH_INTERVAL_MIN; |
| 3340 | else if (logSwitchInterval > LOG_SWITCH_INTERVAL_MAX) |
| 3341 | logSwitchInterval = LOG_SWITCH_INTERVAL_MAX; |
| 3342 | } |
| 3343 | |
| 3344 | /* checkpoint interval */ |
| 3345 | if (cpInterval != null) |
| 3346 | { |
| 3347 | checkpointInterval = Integer.parseInt(cpInterval); |
| 3348 | if (checkpointInterval < CHECKPOINT_INTERVAL_MIN) |
| 3349 | checkpointInterval = CHECKPOINT_INTERVAL_MIN; |
| 3350 | else if(checkpointInterval > CHECKPOINT_INTERVAL_MAX) |
| 3351 | checkpointInterval = CHECKPOINT_INTERVAL_MAX; |
| 3352 | } |
| 3353 | } |
| 3354 | |
| 3355 | public int performWork(ContextManager context) |
| 3356 | { |
| 3357 | synchronized(this) |
| 3358 | { |
| 3359 | if (corrupt != null) |
| 3360 | return Serviceable.DONE; // don't do this again. |
| 3361 | } |
| 3362 | |
| 3363 | // check to see if checkpointInterval and logSwitchInterval has changed |
| 3364 | AccessFactory af = |
| 3365 | (AccessFactory)Monitor.getServiceModule(this, AccessFactory.MODULE); |
| 3366 | |
| 3367 | try |
| 3368 | { |
| 3369 | if (af != null) |
| 3370 | { |
| 3371 | TransactionController tc = null; |
| 3372 | try |
| 3373 | { |
| 3374 | tc = af.getAndNameTransaction( |
| 3375 | context, AccessFactoryGlobals.SYS_TRANS_NAME); |
| 3376 | |
| 3377 | getLogFactoryProperties(tc); |
| 3378 | } |
| 3379 | finally |
| 3380 | { |
| 3381 | if (tc != null) |
| 3382 | tc.commit(); |
| 3383 | } |
| 3384 | } |
| 3385 | |
| 3386 | // checkpoint will start its own internal transaction on the current |
| 3387 | // context. |
| 3388 | rawStoreFactory.checkpoint(); |
| 3389 | } |
| 3390 | catch (StandardException se) |
| 3391 | { |
| 3392 | Monitor.logTextMessage(MessageId.LOG_CHECKPOINT_EXCEPTION); |
| 3393 | logErrMsg(se); |
| 3394 | } |
| 3395 | catch (ShutdownException shutdown) |
| 3396 | { |
| 3397 | // If we are shutting down, just ignore the error and let the |
| 3398 | // system go down without printing errors to the log. |
| 3399 | } |
| 3400 | |
| 3401 | checkpointDaemonCalled = false; |
| 3402 | |
| 3403 | return Serviceable.DONE; |
| 3404 | } |
| 3405 | |
| 3406 | |
| 3407 | /* |
| 3408 | ** Implementation specific methods |
| 3409 | */ |
| 3410 | |
| 3411 | /** |
| 3412 | Append length bytes of data to the log prepended by a long log instant |
| 3413 | and followed by 4 bytes of length information. |
| 3414 | |
| 3415 | <P> |
| 3416 | This method is synchronized to ensure log records are added sequentially |
| 3417 | to the end of the log. |
| 3418 | |
| 3419 | <P>MT- single threaded through this log factory. Log records are |
| 3420 | appended one at a time. |
| 3421 | |
| 3422 | @exception StandardException Log Full. |
| 3423 | |
| 3424 | */ |
| 3425 | protected long appendLogRecord(byte[] data, int offset, int length, |
| 3426 | byte[] optionalData, int optionalDataOffset, int optionalDataLength) |
| 3427 | throws StandardException |
| 3428 | { |
| 3429 | long instant; |
| 3430 | boolean testIncompleteLogWrite = false; |
| 3431 | |
| 3432 | if (ReadOnlyDB) |
| 3433 | { |
| 3434 | throw StandardException.newException( |
| 3435 | SQLState.LOG_READ_ONLY_DB_UPDATE); |
| 3436 | } |
| 3437 | |
| 3438 | if (length <= 0) |
| 3439 | { |
| 3440 | throw StandardException.newException( |
| 3441 | SQLState.LOG_ZERO_LENGTH_LOG_RECORD); |
| 3442 | } |
| 3443 | |
| 3444 | // RESOLVE: calculate checksum here |
| 3445 | if (SanityManager.DEBUG) |
| 3446 | { |
| 3447 | if (SanityManager.DEBUG_ON(TEST_LOG_INCOMPLETE_LOG_WRITE)) |
| 3448 | { |
| 3449 | /// /// /// /// /// /// /// /// /// /// |
| 3450 | // |
| 3451 | // go into this alternate route instead |
| 3452 | // |
| 3453 | /// /// /// /// /// /// /// /// /// /// |
| 3454 | return logtest_appendPartialLogRecord(data, offset, length, |
| 3455 | optionalData, |
| 3456 | optionalDataOffset, |
| 3457 | optionalDataLength); |
| 3458 | |
| 3459 | } |
| 3460 | |
| 3461 | } |
| 3462 | |
| 3463 | try |
| 3464 | { |
| 3465 | if (SanityManager.DEBUG) |
| 3466 | { |
| 3467 | if (SanityManager.DEBUG_ON(TEST_LOG_FULL)) |
| 3468 | testLogFull(); // if log is 'full' this routine will throw an |
| 3469 | // exception |
| 3470 | } |
| 3471 | |
| 3472 | synchronized (this) |
| 3473 | { |
| 3474 | // has someone else found a problem in the raw store? |
| 3475 | if (corrupt != null) |
| 3476 | { |
| 3477 | throw StandardException.newException( |
| 3478 | SQLState.LOG_STORE_CORRUPT, corrupt); |
| 3479 | } |
| 3480 | |
| 3481 | if (logOut == null) |
| 3482 | { |
| 3483 | throw StandardException.newException(SQLState.LOG_NULL); |
| 3484 | } |
| 3485 | |
| 3486 | /* |
| 3487 | * NOTE!! |
| 3488 | * |
| 3489 | * subclass which logs special record to the stream depends on |
| 3490 | * the EXACT byte sequence of the following segment of code. |
| 3491 | * If you change this, not only will you need to write upgrade |
| 3492 | * code for this class, you also need to find all the subclass |
| 3493 | * which write log record to log stream directly to make sure |
| 3494 | * they are OK |
| 3495 | */ |
| 3496 | |
| 3497 | // see if the log file is too big, if it is, switch to the next |
| 3498 | // log file |
| 3499 | if ((endPosition + LOG_RECORD_OVERHEAD + length) >= |
| 3500 | LogCounter.MAX_LOGFILE_SIZE) |
| 3501 | { |
| 3502 | switchLogFile(); |
| 3503 | |
| 3504 | // still too big?? Giant log record? |
| 3505 | if ((endPosition + LOG_RECORD_OVERHEAD + length) >= |
| 3506 | LogCounter.MAX_LOGFILE_SIZE) |
| 3507 | { |
| 3508 | throw StandardException.newException( |
| 3509 | SQLState.LOG_EXCEED_MAX_LOG_FILE_SIZE, |
| 3510 | new Long(logFileNumber), |
| 3511 | new Long(endPosition), |
| 3512 | new Long(length), |
| 3513 | new Long(LogCounter.MAX_LOGFILE_SIZE)); |
| 3514 | } |
| 3515 | } |
| 3516 | |
| 3517 | //reserve the space for the checksum log record |
| 3518 | endPosition += logOut.reserveSpaceForChecksum(length, logFileNumber,endPosition); |
| 3519 | |
| 3520 | // don't call currentInstant since we are already in a |
| 3521 | // synchronzied block |
| 3522 | instant = |
| 3523 | LogCounter.makeLogInstantAsLong(logFileNumber, endPosition); |
| 3524 | |
| 3525 | logOut.writeLogRecord( |
| 3526 | length, instant, data, offset, |
| 3527 | optionalData, optionalDataOffset, optionalDataLength); |
| 3528 | |
| 3529 | if (optionalDataLength != 0) |
| 3530 | { |
| 3531 | if (SanityManager.DEBUG) |
| 3532 | { |
| 3533 | if (optionalData == null) |
| 3534 | SanityManager.THROWASSERT( |
| 3535 | "optionalDataLength = " + optionalDataLength + |
| 3536 | " with null Optional data"); |
| 3537 | |
| 3538 | if (optionalData.length < |
| 3539 | (optionalDataOffset+optionalDataLength)) |
| 3540 | SanityManager.THROWASSERT( |
| 3541 | "optionalDataLength = " + optionalDataLength + |
| 3542 | " optionalDataOffset = " + optionalDataOffset + |
| 3543 | " optionalData.length = " + optionalData.length); |
| 3544 | } |
| 3545 | } |
| 3546 | |
| 3547 | endPosition += (length + LOG_RECORD_OVERHEAD); |
| 3548 | } |
| 3549 | } |
| 3550 | catch (IOException ioe) |
| 3551 | { |
| 3552 | throw markCorrupt(StandardException.newException( |
| 3553 | SQLState.LOG_FULL, ioe)); |
| 3554 | } |
| 3555 | |
| 3556 | return instant; |
| 3557 | } |
| 3558 | |
| 3559 | /* |
| 3560 | * Misc private functions to access the log |
| 3561 | */ |
| 3562 | |
| 3563 | /** |
| 3564 | Get the current log instant - this is the log instant of the Next log |
| 3565 | record to be written out |
| 3566 | <P> MT - This method is synchronized to ensure that it always points to |
| 3567 | the end of a log record, not the middle of one. |
| 3568 | */ |
| 3569 | protected synchronized long currentInstant() |
| 3570 | { |
| 3571 | return LogCounter.makeLogInstantAsLong(logFileNumber, endPosition); |
| 3572 | } |
| 3573 | |
| 3574 | protected synchronized long endPosition() |
| 3575 | { |
| 3576 | return endPosition; |
| 3577 | } |
| 3578 | |
| 3579 | /** |
| 3580 | Return the current log file number. |
| 3581 | |
| 3582 | <P> MT - this method is synchronized so that |
| 3583 | it is not in the middle of being changed by swithLogFile |
| 3584 | */ |
| 3585 | private synchronized long getLogFileNumber() |
| 3586 | { |
| 3587 | return logFileNumber; |
| 3588 | } |
| 3589 | |
| 3590 | /** |
| 3591 | Get the first valid log instant - this is the beginning of the first |
| 3592 | log file |
| 3593 | |
| 3594 | <P>MT- synchronized on this |
| 3595 | */ |
| 3596 | private synchronized long firstLogInstant() |
| 3597 | { |
| 3598 | return LogCounter.makeLogInstantAsLong(firstLogFileNumber, LOG_FILE_HEADER_SIZE); |
| 3599 | } |
| 3600 | |
| 3601 | /** |
| 3602 | Flush the log such that the log record written with the instant |
| 3603 | wherePosition is guaranteed to be on disk. |
| 3604 | |
| 3605 | <P>MT - only one flush is allowed to be taking place at any given time |
| 3606 | (RESOLVE: right now it single thread thru the log factory while the log |
| 3607 | is frozen) |
| 3608 | |
| 3609 | @exception StandardException cannot sync log file |
| 3610 | |
| 3611 | */ |
| 3612 | protected void flush(long fileNumber, long wherePosition) throws StandardException |
| 3613 | { |
| 3614 | |
| 3615 | long potentialLastFlush = 0; |
| 3616 | |
| 3617 | synchronized (this) |
| 3618 | { |
| 3619 | if (Performance.MEASURE) |
| 3620 | mon_flushCalls++; |
| 3621 | try |
| 3622 | { |
| 3623 | boolean waited; |
| 3624 | do |
| 3625 | { |
| 3626 | // THIS CORRUPT CHECK MUST BE FIRST, before any check that |
| 3627 | // sees if the log has already been flushed to this |
| 3628 | // point. This is based upon the assumption that every |
| 3629 | // dirty page in the cache must call flush() before it is |
| 3630 | // written out. has someone else found a problem in the |
| 3631 | // raw store? |
| 3632 | |
| 3633 | if (corrupt != null) |
| 3634 | { |
| 3635 | throw StandardException.newException( |
| 3636 | SQLState.LOG_STORE_CORRUPT, corrupt); |
| 3637 | } |
| 3638 | |
| 3639 | // now check if database is frozen |
| 3640 | while (isFrozen) |
| 3641 | { |
| 3642 | try |
| 3643 | { |
| 3644 | wait(); |
| 3645 | } |
| 3646 | catch (InterruptedException ie) |
| 3647 | { |
| 3648 | throw StandardException.interrupt(ie); |
| 3649 | } |
| 3650 | } |
| 3651 | |
| 3652 | // if we are just testing to see to see the database is |
| 3653 | // frozen or corrupt (wherePosition == INVALID_LOG_INSTANT) |
| 3654 | // then we can return now. |
| 3655 | // if the log file is already flushed up to where we are |
| 3656 | // interested in, just return. |
| 3657 | if (wherePosition == LogCounter.INVALID_LOG_INSTANT || |
| 3658 | fileNumber < logFileNumber || |
| 3659 | wherePosition < lastFlush) |
| 3660 | { |
| 3661 | return; |
| 3662 | } |
| 3663 | |
| 3664 | // if we are not corrupt and we are in the middle of redo, |
| 3665 | // we know the log record has already been flushed since we haven't written any log |
| 3666 | // yet. |
| 3667 | if (recoveryNeeded && inRedo) |
| 3668 | { |
| 3669 | return; |
| 3670 | } |
| 3671 | |
| 3672 | |
| 3673 | if (SanityManager.DEBUG) |
| 3674 | { |
| 3675 | if (fileNumber > getLogFileNumber()) |
| 3676 | SanityManager.THROWASSERT( |
| 3677 | "trying to flush a file that is not there yet " + |
| 3678 | fileNumber + " " + logFileNumber); |
| 3679 | |
| 3680 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 3681 | { |
| 3682 | SanityManager.DEBUG( |
| 3683 | DBG_FLAG, "Flush log to " + wherePosition); |
| 3684 | } |
| 3685 | } |
| 3686 | |
| 3687 | // There could be multiple threads who wants to flush the |
| 3688 | // log file, see if I can be the one. |
| 3689 | if (logBeingFlushed) |
| 3690 | { |
| 3691 | waited = true; |
| 3692 | try |
| 3693 | { |
| 3694 | if (Performance.MEASURE) |
| 3695 | mon_numLogFlushWaits++; |
| 3696 | wait(); // release log semaphore to let non-flushing |
| 3697 | // threads log stuff while all the flushing |
| 3698 | // threads wait. |
| 3699 | |
| 3700 | // now we continue back to see if the sync |
| 3701 | // we waited for, flushed the portion |
| 3702 | // of log we are interested in. |
| 3703 | } |
| 3704 | catch (InterruptedException ie) |
| 3705 | { |
| 3706 | throw StandardException.interrupt(ie); |
| 3707 | } |
| 3708 | } |
| 3709 | else |
| 3710 | { |
| 3711 | waited = false; |
| 3712 | |
| 3713 | // logBeingFlushed is false, I am flushing the log now. |
| 3714 | if(!isWriteSynced) |
| 3715 | { |
| 3716 | // Flush any data from the buffered log |
| 3717 | logOut.flushLogAccessFile(); |
| 3718 | }else |
| 3719 | { |
| 3720 | //add active buffers to dirty buffer list |
| 3721 | //to flush to the disk. |
| 3722 | logOut.switchLogBuffer(); |
| 3723 | } |
| 3724 | |
| 3725 | potentialLastFlush = endPosition; // we will flush to to the end |
| 3726 | |
| 3727 | // once logBeingFlushed is set, need to release |
| 3728 | // the logBeingFlushed flag in finally block. |
| 3729 | logBeingFlushed = true; |
| 3730 | } |
| 3731 | |
| 3732 | } while (waited) ; |
| 3733 | // if I have waited, go down do loop again - hopefully, |
| 3734 | // someone else have already flushed it for me already. |
| 3735 | } |
| 3736 | catch (IOException ioe) |
| 3737 | { |
| 3738 | throw markCorrupt(StandardException.newException( |
| 3739 | SQLState.LOG_CANNOT_FLUSH, |
| 3740 | ioe, |
| 3741 | getLogFileName(logFileNumber).getPath())); |
| 3742 | } |
| 3743 | } // unfreeze log manager to accept more log records |
| 3744 | |
| 3745 | boolean syncSuceed = false; |
| 3746 | try |
| 3747 | { |
| 3748 | if (SanityManager.DEBUG) |
| 3749 | { |
| 3750 | SanityManager.ASSERT(logBeingFlushed, |
| 3751 | "flushing log without logBeingFlushed set"); |
| 3752 | SanityManager.ASSERT(potentialLastFlush > 0, |
| 3753 | "potentialLastFlush not set"); |
| 3754 | |
| 3755 | if (SanityManager.DEBUG_ON(TEST_LOG_FULL)) |
| 3756 | testLogFull(); |
| 3757 | |
| 3758 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 3759 | SanityManager.DEBUG(DBG_FLAG, "Begin log sync..."); |
| 3760 | } |
| 3761 | |
| 3762 | |
| 3763 | if (Performance.MEASURE) |
| 3764 | mon_syncCalls++; |
| 3765 | |
| 3766 | if (isWriteSynced) |
| 3767 | { |
| 3768 | //LogAccessFile.flushDirtyBuffers() will allow only one write |
| 3769 | //sync at a time, flush requests will get queued |
| 3770 | logOut.flushDirtyBuffers(); |
| 3771 | } |
| 3772 | else |
| 3773 | { |
| 3774 | if (!logNotSynced) |
| 3775 | logOut.syncLogAccessFile(); |
| 3776 | } |
| 3777 | |
| 3778 | syncSuceed = true; |
| 3779 | |
| 3780 | if (SanityManager.DEBUG) |
| 3781 | { |
| 3782 | if (SanityManager.DEBUG_ON(LogToFile.DBG_FLAG)) |
| 3783 | SanityManager.DEBUG(DBG_FLAG, "end log sync."); |
| 3784 | } |
| 3785 | } |
| 3786 | catch (SyncFailedException sfe) |
| 3787 | { |
| 3788 | throw markCorrupt(StandardException.newException( |
| 3789 | SQLState.LOG_CANNOT_FLUSH, |
| 3790 | sfe, |
| 3791 | getLogFileName(logFileNumber).getPath())); |
| 3792 | } |
| 3793 | catch (IOException ioe) |
| 3794 | { |
| 3795 | throw markCorrupt(StandardException.newException( |
| 3796 | SQLState.LOG_CANNOT_FLUSH, |
| 3797 | ioe, |
| 3798 | getLogFileName(logFileNumber).getPath())); |
| 3799 | } |
| 3800 | finally |
| 3801 | { |
| 3802 | synchronized(this) |
| 3803 | { |
| 3804 | logBeingFlushed = false; // done flushing |
| 3805 | |
| 3806 | // update lastFlush under synchronized this instead of synchronized(logOut) |
| 3807 | if (syncSuceed) |
| 3808 | { |
| 3809 | lastFlush = potentialLastFlush; |
| 3810 | } |
| 3811 | |
| 3812 | |
| 3813 | // We may actually have flushed more than that because someone |
| 3814 | // may have done a logOut.flushBuffer right before the sync |
| 3815 | // call. But this is guarenteed to be flushed. |
| 3816 | notifyAll(); |
| 3817 | } |
| 3818 | } |
| 3819 | |
| 3820 | |
| 3821 | // get checkpoint Daemon to work |
| 3822 | if ((logWrittenFromLastCheckPoint + potentialLastFlush) > checkpointInterval && |
| 3823 | checkpointDaemon != null && !checkpointDaemonCalled && !inLogSwitch) |
| 3824 | { |
| 3825 | // following synchronized block is required to make |
| 3826 | // sure only one checkpoint request get scheduled. |
| 3827 | synchronized(this) |
| 3828 | { |
| 3829 | // recheck if checkpoint is still required, it is possible some other |
| 3830 | // thread might have already scheduled a checkpoint and completed it. |
| 3831 | if ((logWrittenFromLastCheckPoint + potentialLastFlush) > checkpointInterval && |
| 3832 | checkpointDaemon != null && !checkpointDaemonCalled && !inLogSwitch) |
| 3833 | { |
| 3834 | checkpointDaemonCalled = true; |
| 3835 | checkpointDaemon.serviceNow(myClientNumber); |
| 3836 | } |
| 3837 | } |
| 3838 | |
| 3839 | }else |
| 3840 | { |
| 3841 | // switch the log if required, this case will occur |
| 3842 | // if log switch interval is less than the checkpoint interval |
| 3843 | // other wise , checkpoint daemon would be doing log switches along |
| 3844 | // with the checkpoints. |
| 3845 | if (potentialLastFlush > logSwitchInterval && |
| 3846 | !checkpointDaemonCalled && !inLogSwitch) |
| 3847 | { |
| 3848 | // following synchronized block is required to make sure only |
| 3849 | // one thread switches the log file at a time. |
| 3850 | synchronized(this) |
| 3851 | { |
| 3852 | // recheck if log switch is still required, it is possible some other |
| 3853 | // thread might have already switched the log file. |
| 3854 | if (potentialLastFlush > logSwitchInterval && |
| 3855 | !checkpointDaemonCalled && !inLogSwitch) |
| 3856 | { |
| 3857 | inLogSwitch = true; |
| 3858 | switchLogFile(); |
| 3859 | } |
| 3860 | } |
| 3861 | } |
| 3862 | } |
| 3863 | } |
| 3864 | |
| 3865 | /** |
| 3866 | * Utility routine to call sync() on the input file descriptor. |
| 3867 | * <p> |
| 3868 | */ |
| 3869 | private void syncFile( StorageRandomAccessFile raf) |
| 3870 | throws StandardException |
| 3871 | { |
| 3872 | for( int i=0; ; ) |
| 3873 | { |
| 3874 | // 3311: JVM sync call sometimes fails under high load against NFS |
| 3875 | // mounted disk. We re-try to do this 20 times. |
| 3876 | try |
| 3877 | { |
| 3878 | raf.sync( false); |
| 3879 | |
| 3880 | // the sync succeed, so return |
| 3881 | break; |
| 3882 | } |
| 3883 | catch (IOException ioe) |
| 3884 | { |
| 3885 | i++; |
| 3886 | try |
| 3887 | { |
| 3888 | // wait for .2 of a second, hopefully I/O is done by now |
| 3889 | // we wait a max of 4 seconds before we give up |
| 3890 | Thread.sleep(200); |
| 3891 | } |
| 3892 | catch( InterruptedException ie ) |
| 3893 | { |
| 3894 | //does not matter weather I get interrupted or not |
| 3895 | } |
| 3896 | |
| 3897 | if( i > 20 ) |
| 3898 | { |
| 3899 | throw StandardException.newException( |
| 3900 | SQLState.LOG_FULL, ioe); |
| 3901 | } |
| 3902 | } |
| 3903 | } |
| 3904 | } |
| 3905 | |
| 3906 | |
| 3907 | /** |
| 3908 | Open a forward scan of the transaction log. |
| 3909 | |
| 3910 | <P> MT- read only |
| 3911 | @exception StandardException Standard cloudscape exception policy |
| 3912 | */ |
| 3913 | public LogScan openForwardsFlushedScan(LogInstant startAt) |
| 3914 | throws StandardException |
| 3915 | { |
| 3916 | checkCorrupt(); |
| 3917 | |
| 3918 | // no need to flush the buffer as it's a flushed scan |
| 3919 | |
| 3920 | return new FlushedScan(this,((LogCounter)startAt).getValueAsLong()); |
| 3921 | } |
| 3922 | |
| 3923 | |
| 3924 | /** |
| 3925 | Get a forwards scan |
| 3926 | |
| 3927 | @exception StandardException Standard Cloudscape error policy |
| 3928 | */ |
| 3929 | public LogScan openForwardsScan(LogInstant startAt,LogInstant stopAt) |
| 3930 | throws StandardException |
| 3931 | { |
| 3932 | try |
| 3933 | { |
| 3934 | long startLong; |
| 3935 | |
| 3936 | if (startAt == null) |
| 3937 | startLong = LogCounter.INVALID_LOG_INSTANT; |
| 3938 | else |
| 3939 | startLong = ((LogCounter)startAt).getValueAsLong(); |
| 3940 | |
| 3941 | return openForwardsScan(startLong, stopAt); |
| 3942 | } |
| 3943 | |
| 3944 | catch (IOException ioe) |
| 3945 | { |
| 3946 | throw markCorrupt(StandardException.newException( |
| 3947 | SQLState.LOG_IO_ERROR, ioe)); |
| 3948 | } |
| 3949 | |
| 3950 | } |
| 3951 | |
| 3952 | public final boolean databaseEncrypted() |
| 3953 | { |
| 3954 | return databaseEncrypted; |
| 3955 | } |
| 3956 | |
| 3957 | |
| 3958 | /* |
| 3959 | * Set that the database is encrypted, all the data in the |
| 3960 | * transaction log should be encrypted. |
| 3961 | */ |
| 3962 | public void setDatabaseEncrypted() |
| 3963 | { |
| 3964 | databaseEncrypted = true; |
| 3965 | } |
| 3966 | |
| 3967 | /* |
| 3968 | * setup log for encryption. |
| 3969 | */ |
| 3970 | public void setupLogEncryption() throws StandardException |
| 3971 | { |
| 3972 | // switch the database to a new log file, so that |
| 3973 | // new encrytion will start on new log file. |
| 3974 | switchLogFile(); |
| 3975 | } |
| 3976 | |
| 3977 | |
| 3978 | /** |
| 3979 | @see RawStoreFactory#encrypt |
| 3980 | @exception StandardException Standard Cloudscape Error Policy |
| 3981 | */ |
| 3982 | public int encrypt(byte[] cleartext, int offset, int length, |
| 3983 | byte[] ciphertext, int outputOffset) |
| 3984 | throws StandardException |
| 3985 | { |
| 3986 | return rawStoreFactory.encrypt(cleartext, offset, length, |
| 3987 | ciphertext, outputOffset, false); |
| 3988 | } |
| 3989 | |
| 3990 | /** |
| 3991 | @see RawStoreFactory#decrypt |
| 3992 | @exception StandardException Standard Cloudscape Error Policy |
| 3993 | */ |
| 3994 | public int decrypt(byte[] ciphertext, int offset, int length, |
| 3995 | byte[] cleartext, int outputOffset) |
| 3996 | throws StandardException |
| 3997 | { |
| 3998 | return rawStoreFactory.decrypt(ciphertext, offset, length, cleartext, outputOffset); |
| 3999 | } |
| 4000 | |
| 4001 | /** |
| 4002 | return the encryption block size used during encrypted db creation |
| 4003 | */ |
| 4004 | public int getEncryptionBlockSize() |
| 4005 | { |
| 4006 | return rawStoreFactory.getEncryptionBlockSize(); |
| 4007 | } |
| 4008 | |
| 4009 | /** |
| 4010 | returns the length that will make the data to be multiple of encryption |
| 4011 | block size based on the given length. Block cipher algorithms like DES |
| 4012 | and Blowfish ..etc require their input to be an exact multiple of the block size. |
| 4013 | */ |
| 4014 | public int getEncryptedDataLength(int length) |
| 4015 | { |
| 4016 | if ((length % getEncryptionBlockSize()) != 0) |
| 4017 | { |
| 4018 | return length + getEncryptionBlockSize() - (length % getEncryptionBlockSize()); |
| 4019 | } |
| 4020 | |
| 4021 | return length; |
| 4022 | } |
| 4023 | |
| 4024 | |
| 4025 | |
| 4026 | /** |
| 4027 | Get the instant of the first record which was not |
| 4028 | flushed. |
| 4029 | |
| 4030 | <P>This only works after running recovery the first time. |
| 4031 | <P>MT - RESOLVE: |
| 4032 | */ |
| 4033 | public synchronized LogInstant getFirstUnflushedInstant() |
| 4034 | { |
| 4035 | if (SanityManager.DEBUG) |
| 4036 | SanityManager.ASSERT(logFileNumber > 0 && lastFlush > 0); |
| 4037 | |
| 4038 | return new LogCounter(logFileNumber,lastFlush); |
| 4039 | } |
| 4040 | |
| 4041 | |
| 4042 | /** |
| 4043 | * Backup restore - stop sending log record to the log stream |
| 4044 | * @exception StandardException Standard Cloudscape error policy |
| 4045 | */ |
| 4046 | public void freezePersistentStore() throws StandardException |
| 4047 | { |
| 4048 | // if I get into this synchronized block, I know I am not in the middle |
| 4049 | // of a write because writing to the log file is synchronized under this. |
| 4050 | synchronized(this) |
| 4051 | { |
| 4052 | isFrozen = true; |
| 4053 | } |
| 4054 | } |
| 4055 | |
| 4056 | /** |
| 4057 | * Backup restore - start sending log record to the log stream |
| 4058 | * @exception StandardException Standard Cloudscape error policy |
| 4059 | */ |
| 4060 | public void unfreezePersistentStore() throws StandardException |
| 4061 | { |
| 4062 | synchronized(this) |
| 4063 | { |
| 4064 | isFrozen = false; |
| 4065 | notifyAll(); |
| 4066 | } |
| 4067 | } |
| 4068 | |
| 4069 | /** |
| 4070 | * Backup restore - is the log being archived to some directory? |
| 4071 | * if log archive mode is enabled return true else false |
| 4072 | */ |
| 4073 | public boolean logArchived() |
| 4074 | { |
| 4075 | return logArchived; |
| 4076 | } |
| 4077 | |
| 4078 | /** |
| 4079 | Check to see if a database has been upgraded to the required |
| 4080 | level in order to use a store feature. |
| 4081 | @param requiredMajorVersion required database Engine major version |
| 4082 | @param requiredMinorVersion required database Engine minor version |
| 4083 | @return True if the database has been upgraded to the required level, false otherwise. |
| 4084 | **/ |
| 4085 | boolean checkVersion(int requiredMajorVersion, int requiredMinorVersion) |
| 4086 | { |
| 4087 | if(onDiskMajorVersion > requiredMajorVersion ) |
| 4088 | { |
| 4089 | return true; |
| 4090 | } |
| 4091 | else |
| 4092 | { |
| 4093 | if(onDiskMajorVersion == requiredMajorVersion && |
| 4094 | onDiskMinorVersion >= requiredMinorVersion) |
| 4095 | return true; |
| 4096 | } |
| 4097 | |
| 4098 | return false; |
| 4099 | } |
| 4100 | |
| 4101 | /* |
| 4102 | ** Sending information to the user without throwing exception. |
| 4103 | ** There are times when unusual external or system related things happen in |
| 4104 | ** the log which the user may be interested in but which doesn't impede on |
| 4105 | ** the normal operation of the store. When such an event occur, just send |
| 4106 | ** a message or a warning message to the user rather than throw an |
| 4107 | ** exception, which will rollback a user transaction or crash the database. |
| 4108 | ** |
| 4109 | ** logErrMsg - sends a warning message to the user |
| 4110 | */ |
| 4111 | |
| 4112 | |
| 4113 | /** |
| 4114 | Print error message to user about the log |
| 4115 | MT - not needed, informational only |
| 4116 | */ |
| 4117 | protected void logErrMsg(String msg) |
| 4118 | { |
| 4119 | logErrMsgForDurabilityTestModeNoSync(); |
| 4120 | Monitor.logTextMessage(MessageId.LOG_BEGIN_ERROR); |
| 4121 | Monitor.logMessage(msg); |
| 4122 | Monitor.logTextMessage(MessageId.LOG_END_ERROR); |
| 4123 | } |
| 4124 | |
| 4125 | /** |
| 4126 | Print error message to user about the log |
| 4127 | MT - not needed, informational only |
| 4128 | */ |
| 4129 | protected void logErrMsg(Throwable t) |
| 4130 | { |
| 4131 | logErrMsgForDurabilityTestModeNoSync(); |
| 4132 | if (corrupt != null) |
| 4133 | { |
| 4134 | Monitor.logTextMessage(MessageId.LOG_BEGIN_CORRUPT_STACK); |
| 4135 | printErrorStack(corrupt); |
| 4136 | Monitor.logTextMessage(MessageId.LOG_END_CORRUPT_STACK); |
| 4137 | } |
| 4138 | |
| 4139 | if (t != corrupt) |
| 4140 | { |
| 4141 | Monitor.logTextMessage(MessageId.LOG_BEGIN_ERROR_STACK); |
| 4142 | printErrorStack(t); |
| 4143 | Monitor.logTextMessage(MessageId.LOG_END_ERROR_STACK); |
| 4144 | } |
| 4145 | } |
| 4146 | |
| 4147 | |
| 4148 | /** |
| 4149 | * In case of boot errors, and if database is either booted |
| 4150 | * with derby.system.durability=test or was previously at any time booted in |
| 4151 | * this mode, mention in the error message that the error is probably |
| 4152 | * because the derby.system.durability was set. |
| 4153 | * Dont want to waste time to resolve issues in such |
| 4154 | * cases |
| 4155 | * <p> |
| 4156 | * MT - not needed, informational only |
| 4157 | */ |
| 4158 | private void logErrMsgForDurabilityTestModeNoSync() |
| 4159 | { |
| 4160 | if (logNotSynced || wasDBInDurabilityTestModeNoSync) |
| 4161 | { |
| 4162 | Monitor.logTextMessage( |
| 4163 | MessageId.LOG_DURABILITY_TESTMODE_NO_SYNC_ERR, |
| 4164 | Property.DURABILITY_PROPERTY, |
| 4165 | Property.DURABILITY_TESTMODE_NO_SYNC); |
| 4166 | } |
| 4167 | } |
| 4168 | |
| 4169 | /** |
| 4170 | * print stack trace from the Throwable including |
| 4171 | * its nested exceptions |
| 4172 | * @param t trace starts from this error |
| 4173 | */ |
| 4174 | private void printErrorStack(Throwable t) |
| 4175 | { |
| 4176 | ErrorStringBuilder esb = |
| 4177 | new ErrorStringBuilder(Monitor.getStream().getHeader()); |
| 4178 | esb.stackTrace(t); |
| 4179 | Monitor.logMessage(esb.get().toString()); |
| 4180 | esb.reset(); |
| 4181 | } |
| 4182 | |
| 4183 | |
| 4184 | /** |
| 4185 | * Testing support |
| 4186 | */ |
| 4187 | /** |
| 4188 | Writes out a partial log record - takes the appendLogRecord. |
| 4189 | Need to shutdown the database before another log record gets written, |
| 4190 | or the database is not recoverable. |
| 4191 | */ |
| 4192 | private long logtest_appendPartialLogRecord(byte[] data, int offset, |
| 4193 | int length, |
| 4194 | byte[] optionalData, |
| 4195 | int optionalDataOffset, |
| 4196 | int optionalDataLength) |
| 4197 | throws StandardException |
| 4198 | { |
| 4199 | if (SanityManager.DEBUG) |
| 4200 | { |
| 4201 | int bytesToWrite = 1; |
| 4202 | |
| 4203 | String TestPartialLogWrite = PropertyUtil.getSystemProperty(TEST_LOG_PARTIAL_LOG_WRITE_NUM_BYTES); |
| 4204 | if (TestPartialLogWrite != null) |
| 4205 | { |
| 4206 | bytesToWrite = Integer.valueOf(TestPartialLogWrite).intValue(); |
| 4207 | } |
| 4208 | |
| 4209 | Monitor.logMessage("TEST_LOG_INCOMPLETE_LOG_WRITE: writing " + bytesToWrite + |
| 4210 | " bytes out of " + length + " + " + LOG_RECORD_OVERHEAD + " log record"); |
| 4211 | |
| 4212 | long instant; |
| 4213 | try |
| 4214 | { |
| 4215 | |
| 4216 | synchronized (this) |
| 4217 | { |
| 4218 | // reserve the space for the checksum log record |
| 4219 | // NOTE: bytesToWrite include the log record overhead. |
| 4220 | endPosition += |
| 4221 | logOut.reserveSpaceForChecksum(((length + LOG_RECORD_OVERHEAD) |
| 4222 | < bytesToWrite ? length : |
| 4223 | (bytesToWrite - LOG_RECORD_OVERHEAD)), |
| 4224 | logFileNumber,endPosition); |
| 4225 | instant = currentInstant(); |
| 4226 | |
| 4227 | //check if the length of the records to be written is |
| 4228 | //actually smaller than the number of bytesToWrite |
| 4229 | if(length + LOG_RECORD_OVERHEAD < bytesToWrite) |
| 4230 | endPosition += (length + LOG_RECORD_OVERHEAD); |
| 4231 | else |
| 4232 | endPosition += bytesToWrite; |
| 4233 | |
| 4234 | while(true) // so we can break out without returning out of |
| 4235 | // sync block... |
| 4236 | { |
| 4237 | if (bytesToWrite < 4) |
| 4238 | { |
| 4239 | int shift = 3; |
| 4240 | while(bytesToWrite-- > 0) |
| 4241 | { |
| 4242 | logOut.write((byte)((length >>> 8*shift) & 0xFF)); |
| 4243 | shift--; |
| 4244 | } |
| 4245 | break; |
| 4246 | } |
| 4247 | |
| 4248 | // the length before the log record |
| 4249 | logOut.writeInt(length); |
| 4250 | bytesToWrite -= 4; |
| 4251 | |
| 4252 | if (bytesToWrite < 8) |
| 4253 | { |
| 4254 | int shift = 7; |
| 4255 | while(bytesToWrite-- > 0) |
| 4256 | { |
| 4257 | logOut.write((byte)((instant >>> 8*shift) & 0xFF)); |
| 4258 | shift--; |
| 4259 | } |
| 4260 | break; |
| 4261 | } |
| 4262 | |
| 4263 | // the log instant |
| 4264 | logOut.writeLong(instant); |
| 4265 | bytesToWrite -= 8; |
| 4266 | |
| 4267 | if (bytesToWrite < length) |
| 4268 | { |
| 4269 | int dataLength = length - optionalDataLength; |
| 4270 | if(bytesToWrite < dataLength) |
| 4271 | logOut.write(data, offset,bytesToWrite); |
| 4272 | else |
| 4273 | { |
| 4274 | logOut.write(data, offset, dataLength); |
| 4275 | bytesToWrite -= dataLength ; |
| 4276 | if(optionalDataLength != 0 && bytesToWrite > 0) |
| 4277 | logOut.write(optionalData, optionalDataOffset, bytesToWrite); |
| 4278 | } |
| 4279 | break; |
| 4280 | } |
| 4281 | |
| 4282 | // the log data |
| 4283 | logOut.write(data, offset, length - optionalDataLength); |
| 4284 | //write optional data |
| 4285 | if(optionalDataLength != 0) |
| 4286 | logOut.write(optionalData, optionalDataOffset, optionalDataLength); |
| 4287 | |
| 4288 | bytesToWrite -= length; |
| 4289 | |
| 4290 | if (bytesToWrite < 4) |
| 4291 | { |
| 4292 | int shift = 3; |
| 4293 | while(bytesToWrite-- > 0) |
| 4294 | { |
| 4295 | logOut.write((byte)((length >>> 8*shift) & 0xFF)); |
| 4296 | shift--; |
| 4297 | } |
| 4298 | break; |
| 4299 | } |
| 4300 | |
| 4301 | // the length after the log record |
| 4302 | logOut.writeInt(length); |
| 4303 | break; |
| 4304 | |
| 4305 | } |
| 4306 | |
| 4307 | // do make sure the partial write gets on disk by sync'ing it |
| 4308 | flush(logFileNumber, endPosition); |
| 4309 | |
| 4310 | } |
| 4311 | |
| 4312 | |
| 4313 | } |
| 4314 | catch (IOException ioe) |
| 4315 | { |
| 4316 | throw StandardException.newException(SQLState.LOG_FULL, ioe); |
| 4317 | } |
| 4318 | |
| 4319 | return instant; |
| 4320 | } |
| 4321 | return 0; |
| 4322 | } |
| 4323 | |
| 4324 | /** |
| 4325 | Simulate a log full condition |
| 4326 | |
| 4327 | if TEST_LOG_FULL is set to true, then the property |
| 4328 | TEST_RECORD_TO_FILL_LOG indicates the number of times this function is |
| 4329 | call before an IOException simulating a log full condition is raised. |
| 4330 | |
| 4331 | If TEST_RECORD_TO_FILL_LOG is not set, it defaults to 100 log record |
| 4332 | */ |
| 4333 | protected void testLogFull() throws IOException |
| 4334 | { |
| 4335 | if (SanityManager.DEBUG) |
| 4336 | { |
| 4337 | if (test_numRecordToFillLog < 0) |
| 4338 | { |
| 4339 | String RecordToFillLog = PropertyUtil.getSystemProperty(TEST_RECORD_TO_FILL_LOG); |
| 4340 | if (RecordToFillLog != null) |
| 4341 | test_numRecordToFillLog = Integer.valueOf(RecordToFillLog).intValue(); |
| 4342 | else |
| 4343 | test_numRecordToFillLog = 100; |
| 4344 | } |
| 4345 | |
| 4346 | if (++test_logWritten > test_numRecordToFillLog) |
| 4347 | throw new IOException("TestLogFull " + test_numRecordToFillLog + |
| 4348 | " written " + test_logWritten); |
| 4349 | |
| 4350 | } |
| 4351 | } |
| 4352 | |
| 4353 | /** |
| 4354 | * Get the log file to Simulate a log corruption |
| 4355 | * FOR UNIT TESTING USAGE ONLY |
| 4356 | */ |
| 4357 | public StorageRandomAccessFile getLogFileToSimulateCorruption(long filenum) throws IOException, StandardException |
| 4358 | { |
| 4359 | if (SanityManager.DEBUG) |
| 4360 | { |
| 4361 | //long filenum = LogCounter.getLogFileNumber(logInstant); |
| 4362 | // long filepos = LogCounter.getLogFilePosition(logInstant); |
| 4363 | StorageFile fileName = getLogFileName(filenum); |
| 4364 | StorageRandomAccessFile log = null; |
| 4365 | return privRandomAccessFile(fileName, "rw"); |
| 4366 | } |
| 4367 | |
| 4368 | return null; |
| 4369 | |
| 4370 | } |
| 4371 | |
| 4372 | |
| 4373 | /********************************************************************* |
| 4374 | * Log Testing |
| 4375 | * |
| 4376 | * Implementations may use these strings to simulate error conditions for |
| 4377 | * testing purposes. |
| 4378 | * |
| 4379 | *********************************************************************/ |
| 4380 | |
| 4381 | /** |
| 4382 | Set to true if we want the checkpoint to only switch the log but not |
| 4383 | actually do the checkpoint |
| 4384 | */ |
| 4385 | public static final String TEST_LOG_SWITCH_LOG = SanityManager.DEBUG ? "TEST_LOG_SWITCH_LOG" : null ; |
| 4386 | |
| 4387 | /** |
| 4388 | Set to true if we want the up comming log record to be only partially |
| 4389 | written. The database is corrupted if not immediately shutdown. |
| 4390 | Set TEST_LOG_PARTIAL_LOG_WRITE_NUM_BYTES to the number of bytes to write |
| 4391 | out, default is 1 byte. |
| 4392 | */ |
| 4393 | public static final String TEST_LOG_INCOMPLETE_LOG_WRITE = SanityManager.DEBUG ? "TEST_LOG_INCOMPLETE_LOG_WRITE" : null; |
| 4394 | |
| 4395 | /** |
| 4396 | Set to the number of bytes we want the next log record to actually write |
| 4397 | out, only used when TEST_LOG_INCOMPLETE_LOG_WRITE is on. Default is 1 |
| 4398 | byte. |
| 4399 | */ |
| 4400 | public static final String TEST_LOG_PARTIAL_LOG_WRITE_NUM_BYTES = SanityManager.DEBUG ? "derbyTesting.unittest.partialLogWrite" : null; |
| 4401 | |
| 4402 | /** |
| 4403 | Set to true if we want to simulate a log full condition |
| 4404 | */ |
| 4405 | public static final String TEST_LOG_FULL = |
| 4406 | SanityManager.DEBUG ? "TEST_LOG_FULL" : null; |
| 4407 | |
| 4408 | /** |
| 4409 | Set to true if we want to simulate a log full condition while switching log |
| 4410 | */ |
| 4411 | public static final String TEST_SWITCH_LOG_FAIL1 = |
| 4412 | SanityManager.DEBUG ? "TEST_SWITCH_LOG_FAIL1" : null; |
| 4413 | public static final String TEST_SWITCH_LOG_FAIL2 = |
| 4414 | SanityManager.DEBUG ? "TEST_SWITCH_LOG_FAIL2" : null; |
| 4415 | |
| 4416 | |
| 4417 | /** |
| 4418 | Set to the number of log record we want to write before the log is |
| 4419 | simulated to be full. |
| 4420 | */ |
| 4421 | public static final String TEST_RECORD_TO_FILL_LOG = |
| 4422 | SanityManager.DEBUG ? "derbyTesting.unittest.recordToFillLog" : null; |
| 4423 | |
| 4424 | /** |
| 4425 | * Set to true if we want to simulate max possible log file number is |
| 4426 | * being used. |
| 4427 | */ |
| 4428 | public static final String TEST_MAX_LOGFILE_NUMBER = |
| 4429 | SanityManager.DEBUG ? "testMaxLogFileNumber" : null; |
| 4430 | |
| 4431 | |
| 4432 | //enable the log archive mode |
| 4433 | public void enableLogArchiveMode() throws StandardException |
| 4434 | { |
| 4435 | |
| 4436 | //if the log archive mode is already enabled; thre is nothing to do |
| 4437 | if(!logArchived) |
| 4438 | { |
| 4439 | logArchived = true; |
| 4440 | AccessFactory af = |
| 4441 | (AccessFactory)Monitor.getServiceModule(this, AccessFactory.MODULE); |
| 4442 | |
| 4443 | if (af != null) |
| 4444 | { |
| 4445 | TransactionController tc = null; |
| 4446 | tc = af.getTransaction( |
| 4447 | ContextService.getFactory().getCurrentContextManager()); |
| 4448 | tc.setProperty(Property.LOG_ARCHIVE_MODE , "true", true); |
| 4449 | } |
| 4450 | } |
| 4451 | } |
| 4452 | |
| 4453 | // disable the log archive mode |
| 4454 | public void disableLogArchiveMode() throws StandardException |
| 4455 | { |
| 4456 | AccessFactory af = |
| 4457 | (AccessFactory)Monitor.getServiceModule(this, AccessFactory.MODULE); |
| 4458 | if (af != null) |
| 4459 | { |
| 4460 | TransactionController tc = null; |
| 4461 | tc = af.getTransaction(ContextService.getFactory().getCurrentContextManager()); |
| 4462 | tc.setProperty(Property.LOG_ARCHIVE_MODE , "false", true); |
| 4463 | } |
| 4464 | logArchived = false; |
| 4465 | } |
| 4466 | |
| 4467 | //delete the online archived log files |
| 4468 | public void deleteOnlineArchivedLogFiles() |
| 4469 | { |
| 4470 | deleteObsoleteLogfiles(); |
| 4471 | } |
| 4472 | |
| 4473 | |
| 4474 | /* |
| 4475 | * start the transaction log backup, transaction log is is required |
| 4476 | * to bring the database to the consistent state on restore. |
| 4477 | |
| 4478 | * All the log files that are created after the backup starts |
| 4479 | * should be kept around until they are copied into the backup, |
| 4480 | * even if there are checkpoints when backup is in progress. |
| 4481 | * |
| 4482 | * copy the log control files to the backup (the checkpoint recorded in the |
| 4483 | * control files is the backup checkpoint), Restore will use the checkpoint |
| 4484 | * info in these control files to perform recovery to bring |
| 4485 | * the database to the consistent state. and find first log file |
| 4486 | * that need to be copied into the backup to bring the database |
| 4487 | * to the consistent state on restore. |
| 4488 | * |
| 4489 | * In the end, existing log files that are needed to recover from the backup |
| 4490 | * checkpoint are copied into the backup, any log that gets generated after |
| 4491 | * this call are copied into the backup after all the all the information |
| 4492 | * in the data containers is written to the backup, when endLogBackup() |
| 4493 | * is called. |
| 4494 | * |
| 4495 | * @param toDir - location where the log files should be copied to. |
| 4496 | * @exception StandardException Standard Derby error policy |
| 4497 | * |
| 4498 | */ |
| 4499 | public void startLogBackup(File toDir) throws StandardException |
| 4500 | { |
| 4501 | |
| 4502 | // synchronization is necessary to make sure NO parallel |
| 4503 | // checkpoint happens when the current checkpoint information |
| 4504 | // is being copied to the backup. |
| 4505 | |
| 4506 | synchronized(this) |
| 4507 | { |
| 4508 | // wait until the thread that is doing the checkpoint completes it. |
| 4509 | while(inCheckpoint) |
| 4510 | { |
| 4511 | try |
| 4512 | { |
| 4513 | wait(); |
| 4514 | } |
| 4515 | catch (InterruptedException ie) |
| 4516 | { |
| 4517 | throw StandardException.interrupt(ie); |
| 4518 | } |
| 4519 | } |
| 4520 | |
| 4521 | backupInProgress = true; |
| 4522 | |
| 4523 | // copy the control files. |
| 4524 | StorageFile fromFile; |
| 4525 | File toFile; |
| 4526 | // copy the log control file |
| 4527 | fromFile = getControlFileName(); |
| 4528 | toFile = new File(toDir,fromFile.getName()); |
| 4529 | if(!privCopyFile(fromFile, toFile)) |
| 4530 | { |
| 4531 | throw StandardException.newException(SQLState.RAWSTORE_ERROR_COPYING_FILE, |
| 4532 | fromFile, toFile); |
| 4533 | } |
| 4534 | |
| 4535 | // copy the log mirror control file |
| 4536 | fromFile = getMirrorControlFileName(); |
| 4537 | toFile = new File(toDir,fromFile.getName()); |
| 4538 | if(!privCopyFile(fromFile, toFile)) |
| 4539 | { |
| 4540 | throw StandardException.newException(SQLState.RAWSTORE_ERROR_COPYING_FILE, |
| 4541 | fromFile, toFile); |
| 4542 | } |
| 4543 | |
| 4544 | // find the first log file number that is active |
| 4545 | logFileToBackup = getFirstLogNeeded(currentCheckpoint); |
| 4546 | } |
| 4547 | |
| 4548 | // copy all the log files that has to go into the backup |
| 4549 | backupLogFiles(toDir, getLogFileNumber()-1); |
| 4550 | } |
| 4551 | |
| 4552 | /* |
| 4553 | * copy the log files into the given backup location |
| 4554 | * @param toDir - location where the log files should be copied to. |
| 4555 | * @param lastLogFileToBackup - last log file that needs to be copied. |
| 4556 | **/ |
| 4557 | private void backupLogFiles(File toDir, long lastLogFileToBackup) |
| 4558 | throws StandardException |
| 4559 | { |
| 4560 | |
| 4561 | while(logFileToBackup <= lastLogFileToBackup) |
| 4562 | { |
| 4563 | StorageFile fromFile = getLogFileName(logFileToBackup); |
| 4564 | File toFile = new File(toDir, fromFile.getName()); |
| 4565 | if(!privCopyFile(fromFile, toFile)) |
| 4566 | { |
| 4567 | throw StandardException.newException(SQLState.RAWSTORE_ERROR_COPYING_FILE, |
| 4568 | fromFile, toFile); |
| 4569 | } |
| 4570 | logFileToBackup++; |
| 4571 | } |
| 4572 | } |
| 4573 | |
| 4574 | /* |
| 4575 | * copy all the log files that has to go into the backup |
| 4576 | * and mark that backup is compeleted. |
| 4577 | * |
| 4578 | * @param toDir - location where the log files should be copied to. |
| 4579 | * @exception StandardException Standard Derby error policy |
| 4580 | */ |
| 4581 | public void endLogBackup(File toDir) throws StandardException |
| 4582 | { |
| 4583 | long lastLogFileToBackup; |
| 4584 | if (logArchived) |
| 4585 | { |
| 4586 | // when the log is being archived for roll-frward recovery |
| 4587 | // we would like to switch to a new log file. |
| 4588 | // otherwise during restore logfile in the backup could |
| 4589 | // overwrite the more uptodate log files in the |
| 4590 | // online log path. And also we would like to mark the end |
| 4591 | // marker for the log file other wise during roll-forward recovery, |
| 4592 | // if we see a log file with fuzzy end , we think that is the |
| 4593 | // end of the recovery. |
| 4594 | switchLogFile(); |
| 4595 | lastLogFileToBackup = getLogFileNumber()-1 ; |
| 4596 | }else |
| 4597 | { |
| 4598 | // for a plain online backup partiall filled up log file is ok, |
| 4599 | // no need to do a log switch. |
| 4600 | lastLogFileToBackup = getLogFileNumber(); |
| 4601 | } |
| 4602 | |
| 4603 | // backup all the log that got generated after the backup started. |
| 4604 | backupLogFiles(toDir, lastLogFileToBackup); |
| 4605 | |
| 4606 | // mark that backup is completed. |
| 4607 | backupInProgress = false; |
| 4608 | } |
| 4609 | |
| 4610 | |
| 4611 | /* |
| 4612 | * backup is not in progress any more, it failed for some reason. |
| 4613 | **/ |
| 4614 | public void abortLogBackup() |
| 4615 | { |
| 4616 | backupInProgress = false; |
| 4617 | } |
| 4618 | |
| 4619 | |
| 4620 | // Is the transaction in rollforward recovery |
| 4621 | public boolean inRFR() |
| 4622 | { |
| 4623 | /* |
| 4624 | *Logging System does not differentiate between the |
| 4625 | *crash-recovery and a rollforward recovery. |
| 4626 | *Except in case of rollforward atttempt on |
| 4627 | *read only databases to check for pending Transaction. |
| 4628 | *(See the comments in recovery() function) |
| 4629 | */ |
| 4630 | |
| 4631 | if(recoveryNeeded) |
| 4632 | { |
| 4633 | boolean readOnly = false; |
| 4634 | try{ |
| 4635 | readOnly = !privCanWrite(getControlFileName()); |
| 4636 | }catch(StandardException se) |
| 4637 | { |
| 4638 | //Exception should never have come here |
| 4639 | //because getControlFileName() is called |
| 4640 | //earlier at boot time, if there were problems |
| 4641 | //it should have showed up earlier. |
| 4642 | //We just ignore this error and hope that |
| 4643 | //datafactory must have market it as read only if that is the case. |
| 4644 | } |
| 4645 | |
| 4646 | readOnly = readOnly || (dataFactory == null ? false :dataFactory.isReadOnly()); |
| 4647 | return !readOnly; |
| 4648 | }else{ |
| 4649 | return false; |
| 4650 | } |
| 4651 | } |
| 4652 | |
| 4653 | /** |
| 4654 | * redo a checkpoint during rollforward recovery |
| 4655 | */ |
| 4656 | public void checkpointInRFR(LogInstant cinstant, long redoLWM, DataFactory df) throws StandardException |
| 4657 | { |
| 4658 | //sync the data |
| 4659 | df.checkpoint(); |
| 4660 | |
| 4661 | //write the log control file; this will make sure that restart of the |
| 4662 | //rollfoward recovery will start this log instant next time instead of |
| 4663 | //from the beginning. |
| 4664 | try{ |
| 4665 | if (!writeControlFile(getControlFileName(), ((LogCounter)cinstant).getValueAsLong())) |
| 4666 | { |
| 4667 | throw StandardException.newException( |
| 4668 | SQLState.LOG_CONTROL_FILE, getControlFileName()); |
| 4669 | } |
| 4670 | } |
| 4671 | catch (IOException ioe) |
| 4672 | { |
| 4673 | throw markCorrupt( |
| 4674 | StandardException.newException(SQLState.LOG_IO_ERROR, ioe)); |
| 4675 | } |
| 4676 | //remove the stub files |
| 4677 | df.removeDroppedContainerFileStubs(new LogCounter(redoLWM)); |
| 4678 | |
| 4679 | } |
| 4680 | |
| 4681 | |
| 4682 | /** |
| 4683 | * |
| 4684 | * This function restores logs based on the following attributes |
| 4685 | * are specified on connection URL: |
| 4686 | * Attribute.CREATE_FROM (Create database from backup if it does not exist) |
| 4687 | * Attribute.RESTORE_FROM (Delete the whole database if it exists and then |
| 4688 | * restore it from backup) |
| 4689 | * Attribute.ROLL_FORWARD_RECOVERY_FROM:(Perform Rollforward Recovery; |
| 4690 | * except for the log directory everthing else is replced by the copy from |
| 4691 | * backup. log files in the backup are copied to the existing online log |
| 4692 | * directory. |
| 4693 | * |
| 4694 | * In cases of RESTORE_FROM whole databases directoy is |
| 4695 | * is removed in Directory.java while restoring service.properties |
| 4696 | * so even the log directory is removed. |
| 4697 | * In case of CREATE_FROM , log directoy will not exist if |
| 4698 | * we came so far bacause it should fail if a database already exists. |
| 4699 | * In case ROLL_FORWARD_RECOVERY_FROM log directotry should not be removed. |
| 4700 | * So only thing that needs to be done here is create a |
| 4701 | * a log directory if it does not exists and copy the |
| 4702 | * log files(including control files) that exists in the backup from which |
| 4703 | * we are are trying to restore the database to the onlie log directory. |
| 4704 | */ |
| 4705 | private boolean restoreLogs(Properties properties) throws StandardException |
| 4706 | { |
| 4707 | |
| 4708 | String backupPath =null; |
| 4709 | boolean isCreateFrom = false; |
| 4710 | boolean isRestoreFrom = false; |
| 4711 | |
| 4712 | //check if the user requested for restore/recovery/create from backup |
| 4713 | backupPath = properties.getProperty(Attribute.CREATE_FROM); |
| 4714 | if (backupPath != null) { |
| 4715 | isCreateFrom = true; |
| 4716 | } else { |
| 4717 | backupPath = properties.getProperty(Attribute.RESTORE_FROM); |
| 4718 | if (backupPath != null) { |
| 4719 | isRestoreFrom = true; |
| 4720 | } else { |
| 4721 | backupPath = properties.getProperty( |
| 4722 | Attribute.ROLL_FORWARD_RECOVERY_FROM); |
| 4723 | // if the backup is not NULL then it is a rollforward recovery. |
| 4724 | } |
| 4725 | } |
| 4726 | |
| 4727 | if(backupPath !=null) |
| 4728 | { |
| 4729 | if(!isCreateFrom){ |
| 4730 | if(logDevice == null){ |
| 4731 | /** |
| 4732 | * In restoreFrom/rollForwardRecoveryFrom mode when no |
| 4733 | * logDevice on URL then the log is restored to the same |
| 4734 | * location where the log was when backup was taken. |
| 4735 | * In createFrom mode behaviour is same as when create=true, |
| 4736 | * i.e unless user specifies the logDevice on URL, log will |
| 4737 | * be copied to the database home dir. |
| 4738 | * Note: LOG_DEVICE_AT_BACKUP will get set if log is not in |
| 4739 | * default location(db home). |
| 4740 | */ |
| 4741 | logDevice = |
| 4742 | properties.getProperty(Property.LOG_DEVICE_AT_BACKUP); |
| 4743 | } |
| 4744 | } |
| 4745 | |
| 4746 | getLogStorageFactory(); |
| 4747 | StorageFile logDir; |
| 4748 | logDir = logStorageFactory.newStorageFile( |
| 4749 | LogFactory.LOG_DIRECTORY_NAME); |
| 4750 | |
| 4751 | //remove the log directory in case of restoreFrom |
| 4752 | //if it exist, this happens if the log device is on seperate |
| 4753 | //location than the db home. |
| 4754 | if (isRestoreFrom && logDevice != null) |
| 4755 | { |
| 4756 | if(!privRemoveDirectory(logDir)) |
| 4757 | { |
| 4758 | //it may be just a file, try deleting it |
| 4759 | if(!privDelete(logDir)) |
| 4760 | { |
| 4761 | throw StandardException.newException( |
| 4762 | SQLState.UNABLE_TO_REMOVE_DATA_DIRECTORY, |
| 4763 | getLogDirPath( logDir)); |
| 4764 | } |
| 4765 | } |
| 4766 | } |
| 4767 | |
| 4768 | // if it is a create/restore from backup, |
| 4769 | // create the log directory. |
| 4770 | if (isCreateFrom || isRestoreFrom) { |
| 4771 | createLogDirectory(); |
| 4772 | } |
| 4773 | |
| 4774 | File backupLogDir = new File(backupPath, LogFactory.LOG_DIRECTORY_NAME); |
| 4775 | String[] logfilelist = privList(backupLogDir); |
| 4776 | if(logfilelist !=null) |
| 4777 | { |
| 4778 | for (int i = 0; i < logfilelist.length; i++) |
| 4779 | { |
| 4780 | File blogFile = new File(backupLogDir, logfilelist[i]); |
| 4781 | StorageFile clogFile = logStorageFactory.newStorageFile(logDir, logfilelist[i]); |
| 4782 | if(!privCopyFile(blogFile , clogFile)) |
| 4783 | { |
| 4784 | throw |
| 4785 | StandardException.newException(SQLState.UNABLE_TO_COPY_LOG_FILE, blogFile, clogFile); |
| 4786 | } |
| 4787 | } |
| 4788 | }else |
| 4789 | { |
| 4790 | throw StandardException.newException(SQLState.LOG_DIRECTORY_NOT_FOUND_IN_BACKUP,backupLogDir); |
| 4791 | } |
| 4792 | //we need to switch the log file after redo while |
| 4793 | //doing recovery from backups, otherwise we will |
| 4794 | //be replacing updated log after a restore withe |
| 4795 | // a log in the backup on next restore. |
| 4796 | logSwitchRequired = true; |
| 4797 | |
| 4798 | // log is restored from backup. |
| 4799 | return true; |
| 4800 | } else { |
| 4801 | // log is not restored from backup. |
| 4802 | return false; |
| 4803 | } |
| 4804 | } |
| 4805 | |
| 4806 | /*preallocate the given log File to the logSwitchInterval size; |
| 4807 | *file is extended by writing zeros after the header till |
| 4808 | *the log file size the set by the user. |
| 4809 | */ |
| 4810 | private void preAllocateNewLogFile(StorageRandomAccessFile log) throws IOException, StandardException |
| 4811 | { |
| 4812 | //preallocate a file by writing zeros into it . |
| 4813 | |
| 4814 | if (SanityManager.DEBUG) |
| 4815 | { |
| 4816 | int currentPostion = (int)log.getFilePointer(); |
| 4817 | SanityManager.ASSERT(currentPostion == LOG_FILE_HEADER_SIZE, |
| 4818 | "New Log File Is not Correctly Initialized"); |
| 4819 | } |
| 4820 | |
| 4821 | int amountToWrite = logSwitchInterval - LOG_FILE_HEADER_SIZE ; |
| 4822 | int bufferSize = logBufferSize * 2; |
| 4823 | byte[] emptyBuffer = new byte[bufferSize]; |
| 4824 | int nWrites = amountToWrite/bufferSize; |
| 4825 | int remainingBytes = amountToWrite % bufferSize; |
| 4826 | |
| 4827 | try{ |
| 4828 | while(nWrites-- > 0) |
| 4829 | log.write(emptyBuffer); |
| 4830 | |
| 4831 | if(remainingBytes !=0) |
| 4832 | log.write(emptyBuffer , 0 ,remainingBytes); |
| 4833 | |
| 4834 | //sync the file |
| 4835 | syncFile(log); |
| 4836 | }catch(IOException ie) |
| 4837 | { |
| 4838 | //ignore io exceptions during preallocations |
| 4839 | //because this more for performance improvements |
| 4840 | //system shoulf work fine even without preallocations. |
| 4841 | |
| 4842 | //RESOLVE: If the exception is because of no |
| 4843 | //space, might be good idea to trigger a checkpoint. |
| 4844 | |
| 4845 | //In debug mode throw the exception |
| 4846 | if (SanityManager.DEBUG) |
| 4847 | { |
| 4848 | throw ie; |
| 4849 | } |
| 4850 | } |
| 4851 | } // end of preAllocateNewLogFile |
| 4852 | |
| 4853 | |
| 4854 | /** |
| 4855 | * open the given log file name for writes; if file can not be |
| 4856 | * be opened in write sync mode then disable the write sync mode and |
| 4857 | * open the file in "rw" mode. |
| 4858 | */ |
| 4859 | private StorageRandomAccessFile openLogFileInWriteMode(StorageFile logFile) throws IOException |
| 4860 | { |
| 4861 | StorageRandomAccessFile log; |
| 4862 | try{ |
| 4863 | log = privRandomAccessFile(logFile, "rws"); |
| 4864 | }catch(FileNotFoundException ex) |
| 4865 | { |
| 4866 | // Normally this exception should never occur. For some reason |
| 4867 | // currently on Mac JVM 1.4.2 FileNotFoundException exception is |
| 4868 | // thrown if a file is opened in "rws" mode and if it already |
| 4869 | // exists. Please refere to Derby-1 for more/ details on this issue. |
| 4870 | // Temporary workaround to avoid this problem is to make the logging |
| 4871 | // system use file sync mechanism. |
| 4872 | |
| 4873 | // disable the write sync and open the file in "rw" mode. |
| 4874 | isWriteSynced = false; |
| 4875 | log = privRandomAccessFile(logFile, "rw"); |
| 4876 | } |
| 4877 | |
| 4878 | return log ; |
| 4879 | } |
| 4880 | |
| 4881 | |
| 4882 | private String getLogDirPath( StorageFile logDir) |
| 4883 | { |
| 4884 | if( logDevice == null) |
| 4885 | return logDir.toString(); |
| 4886 | return logDevice + logStorageFactory.getSeparator() + logDir.toString(); |
| 4887 | } // end of getLogDirPath |
| 4888 | |
| 4889 | /* |
| 4890 | Following methods require Priv Blocks to run under a security manager. |
| 4891 | */ |
| 4892 | private int action; |
| 4893 | private StorageFile activeFile; |
| 4894 | private File toFile; |
| 4895 | private String activePerms; |
| 4896 | |
| 4897 | protected boolean privExists(StorageFile file) |
| 4898 | { |
| 4899 | return runBooleanAction(0, file); |
| 4900 | } |
| 4901 | |
| 4902 | protected boolean privDelete(StorageFile file) |
| 4903 | { |
| 4904 | return runBooleanAction(1, file); |
| 4905 | } |
| 4906 | |
| 4907 | private synchronized StorageRandomAccessFile privRandomAccessFile(StorageFile file, String perms) |
| 4908 | throws IOException |
| 4909 | { |
| 4910 | action = 2; |
| 4911 | activeFile = file; |
| 4912 | activePerms = perms; |
| 4913 | try |
| 4914 | { |
| 4915 | return (StorageRandomAccessFile) java.security.AccessController.doPrivileged(this); |
| 4916 | } |
| 4917 | catch (java.security.PrivilegedActionException pae) |
| 4918 | { |
| 4919 | throw (IOException) pae.getException(); |
| 4920 | } |
| 4921 | } |
| 4922 | |
| 4923 | protected boolean privCanWrite(StorageFile file) |
| 4924 | { |
| 4925 | return runBooleanAction(3, file); |
| 4926 | } |
| 4927 | |
| 4928 | protected boolean privMkdirs(StorageFile file) |
| 4929 | { |
| 4930 | return runBooleanAction(4, file); |
| 4931 | } |
| 4932 | |
| 4933 | private synchronized String[] privList(File file) |
| 4934 | { |
| 4935 | action = 8; |
| 4936 | toFile = file; |
| 4937 | |
| 4938 | try |
| 4939 | { |
| 4940 | return (String[]) java.security.AccessController.doPrivileged(this); |
| 4941 | } |
| 4942 | catch (java.security.PrivilegedActionException pae) |
| 4943 | { |
| 4944 | return null; |
| 4945 | } |
| 4946 | } |
| 4947 | |
| 4948 | private synchronized String[] privList(StorageFile file) |
| 4949 | { |
| 4950 | action = 5; |
| 4951 | activeFile = file; |
| 4952 | |
| 4953 | try |
| 4954 | { |
| 4955 | return (String[]) java.security.AccessController.doPrivileged(this); |
| 4956 | } |
| 4957 | catch (java.security.PrivilegedActionException pae) |
| 4958 | { |
| 4959 | return null; |
| 4960 | } |
| 4961 | } |
| 4962 | |
| 4963 | |
| 4964 | private synchronized boolean privCopyFile(StorageFile from, File to) |
| 4965 | { |
| 4966 | action = 6; |
| 4967 | activeFile = from; |
| 4968 | toFile = to; |
| 4969 | try |
| 4970 | { |
| 4971 | return ((Boolean) java.security.AccessController.doPrivileged(this)).booleanValue(); |
| 4972 | } |
| 4973 | catch (java.security.PrivilegedActionException pae) |
| 4974 | { |
| 4975 | return false; |
| 4976 | } |
| 4977 | } |
| 4978 | |
| 4979 | private synchronized boolean privCopyFile(File from, StorageFile to) |
| 4980 | { |
| 4981 | action = 9; |
| 4982 | activeFile = to; |
| 4983 | toFile = from; |
| 4984 | try |
| 4985 | { |
| 4986 | return ((Boolean) java.security.AccessController.doPrivileged(this)).booleanValue(); |
| 4987 | } |
| 4988 | catch (java.security.PrivilegedActionException pae) |
| 4989 | { |
| 4990 | return false; |
| 4991 | } |
| 4992 | } |
| 4993 | |
| 4994 | private boolean privRemoveDirectory(StorageFile file) |
| 4995 | { |
| 4996 | return runBooleanAction(7, file); |
| 4997 | } |
| 4998 | |
| 4999 | |
| 5000 | private synchronized boolean runBooleanAction(int action, StorageFile file) { |
| 5001 | this.action = action; |
| 5002 | this.activeFile = file; |
| 5003 | |
| 5004 | try { |
| 5005 | return ((Boolean) java.security.AccessController.doPrivileged(this)).booleanValue(); |
| 5006 | } catch (java.security.PrivilegedActionException pae) { |
| 5007 | return false; |
| 5008 | } |
| 5009 | } |
| 5010 | |
| 5011 | |
| 5012 | |
| 5013 | |
| 5014 | public final Object run() throws IOException { |
| 5015 | switch (action) { |
| 5016 | case 0: |
| 5017 | // SECURITY PERMISSION - MP1 |
| 5018 | return ReuseFactory.getBoolean(activeFile.exists()); |
| 5019 | case 1: |
| 5020 | // SECURITY PERMISSION - OP5 |
| 5021 | return ReuseFactory.getBoolean(activeFile.delete()); |
| 5022 | case 2: |
| 5023 | // SECURITY PERMISSION - MP1 and/or OP4 |
| 5024 | // dependening on the value of activePerms |
| 5025 | return activeFile.getRandomAccessFile(activePerms); |
| 5026 | case 3: |
| 5027 | // SECURITY PERMISSION - OP4 |
| 5028 | return ReuseFactory.getBoolean(activeFile.canWrite()); |
| 5029 | case 4: |
| 5030 | // SECURITY PERMISSION - OP4 |
| 5031 | return ReuseFactory.getBoolean(activeFile.mkdirs()); |
| 5032 | case 5: |
| 5033 | // SECURITY PERMISSION - MP1 |
| 5034 | return activeFile.list(); |
| 5035 | case 6: |
| 5036 | // SECURITY PERMISSION - OP4 (Have to check these codes ??) |
| 5037 | return ReuseFactory.getBoolean(FileUtil.copyFile(logStorageFactory, activeFile, toFile)); |
| 5038 | case 7: |
| 5039 | // SECURITY PERMISSION - OP4 |
| 5040 | if( ! activeFile.exists()) |
| 5041 | return ReuseFactory.getBoolean( true); |
| 5042 | return ReuseFactory.getBoolean(activeFile.deleteAll()); |
| 5043 | case 8: |
| 5044 | return toFile.list(); |
| 5045 | case 9: |
| 5046 | return ReuseFactory.getBoolean(FileUtil.copyFile( logStorageFactory, toFile, activeFile)); |
| 5047 | |
| 5048 | default: |
| 5049 | return null; |
| 5050 | } |
| 5051 | } |
| 5052 | } |