Re: Backup history file should be replicated in Streaming Replication? - Mailing list pgsql-hackers
From | Heikki Linnakangas |
---|---|
Subject | Re: Backup history file should be replicated in Streaming Replication? |
Date | |
Msg-id | 4B3B54FB.2010601@enterprisedb.com Whole thread Raw |
In response to | Re: Backup history file should be replicated in Streaming Replication? (Fujii Masao <masao.fujii@gmail.com>) |
Responses |
Re: Backup history file should be replicated in Streaming
Replication?
|
List | pgsql-hackers |
Fujii Masao wrote: > On Thu, Dec 24, 2009 at 1:39 PM, Fujii Masao <masao.fujii@gmail.com> wrote: >> On Wed, Dec 23, 2009 at 7:50 PM, Heikki Linnakangas >> <heikki.linnakangas@enterprisedb.com> wrote: >>> Ok. How about writing the history file in pg_stop_backup() for >>> informational purposes only. Ie. never read it, but rely on the WAL >>> records instead. >> Sounds good. I'll make such change as a self-contained patch. > > Done. Please see the attached patch. > > Design: > > * pg_stop_backup writes the backup-end xlog record which contains > the backup starting point. > > * In archive recovery, the startup process doesn't mark the database > as consistent until it has read the backup-end record. > > * A backup history file is still created as in the past, but is never > used. As the patch stands, reachedBackupEnd is never set to true if starting from a restore point after the end-of-backup. We'll need to store the information that we've reached end-of-backup somewhere on disk. Here's is modified patch that adds a new backupStartPoint field to pg_control for that + some other minor editorialization. -- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 783725c..63884eb 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -515,8 +515,7 @@ static void xlog_outrec(StringInfo buf, XLogRecord *record); #endif static void issue_xlog_fsync(void); static void pg_start_backup_callback(int code, Datum arg); -static bool read_backup_label(XLogRecPtr *checkPointLoc, - XLogRecPtr *minRecoveryLoc); +static bool read_backup_label(XLogRecPtr *checkPointLoc); static void rm_redo_error_callback(void *arg); static int get_sync_bit(int method); @@ -5355,7 +5354,6 @@ StartupXLOG(void) bool haveBackupLabel = false; XLogRecPtr RecPtr, checkPointLoc, - backupStopLoc, EndOfLog; uint32 endLogId; uint32 endLogSeg; @@ -5454,7 +5452,7 @@ StartupXLOG(void) recoveryTargetTLI, ControlFile->checkPointCopy.ThisTimeLineID))); - if (read_backup_label(&checkPointLoc, &backupStopLoc)) + if (read_backup_label(&checkPointLoc)) { /* * When a backup_label file is present, we want to roll forward from @@ -5597,11 +5595,23 @@ StartupXLOG(void) ControlFile->prevCheckPoint = ControlFile->checkPoint; ControlFile->checkPoint = checkPointLoc; ControlFile->checkPointCopy = checkPoint; - if (backupStopLoc.xlogid != 0 || backupStopLoc.xrecoff != 0) + if (InArchiveRecovery) + { + /* initialize minRecoveryPoint if not set yet */ + if (XLByteLT(ControlFile->minRecoveryPoint, checkPoint.redo)) + ControlFile->minRecoveryPoint = checkPoint.redo; + } + else { - if (XLByteLT(ControlFile->minRecoveryPoint, backupStopLoc)) - ControlFile->minRecoveryPoint = backupStopLoc; + XLogRecPtr InvalidXLogRecPtr = {0, 0}; + ControlFile->minRecoveryPoint = InvalidXLogRecPtr; } + /* + * set backupStartupPoint if we're starting archive recovery from a + * base backup. + */ + if (haveBackupLabel) + ControlFile->backupStartPoint = checkPoint.redo; ControlFile->time = (pg_time_t) time(NULL); /* No need to hold ControlFileLock yet, we aren't up far enough */ UpdateControlFile(); @@ -5703,15 +5713,9 @@ StartupXLOG(void) InRedo = true; - if (minRecoveryPoint.xlogid == 0 && minRecoveryPoint.xrecoff == 0) - ereport(LOG, - (errmsg("redo starts at %X/%X", - ReadRecPtr.xlogid, ReadRecPtr.xrecoff))); - else - ereport(LOG, - (errmsg("redo starts at %X/%X, consistency will be reached at %X/%X", - ReadRecPtr.xlogid, ReadRecPtr.xrecoff, - minRecoveryPoint.xlogid, minRecoveryPoint.xrecoff))); + ereport(LOG, + (errmsg("redo starts at %X/%X", + ReadRecPtr.xlogid, ReadRecPtr.xrecoff))); /* * Let postmaster know we've started redo now, so that it can @@ -5771,7 +5775,8 @@ StartupXLOG(void) * Have we passed our safe starting point? */ if (!reachedMinRecoveryPoint && - XLByteLE(minRecoveryPoint, EndRecPtr)) + XLByteLE(minRecoveryPoint, EndRecPtr) && + XLogRecPtrIsInvalid(ControlFile->backupStartPoint)) { reachedMinRecoveryPoint = true; ereport(LOG, @@ -5877,7 +5882,9 @@ StartupXLOG(void) * be further ahead --- ControlFile->minRecoveryPoint cannot have been * advanced beyond the WAL we processed. */ - if (InRecovery && XLByteLT(EndOfLog, minRecoveryPoint)) + if (InArchiveRecovery && + (XLByteLT(EndOfLog, minRecoveryPoint) || + !XLogRecPtrIsInvalid(ControlFile->backupStartPoint))) { if (reachedStopPoint) /* stopped because of stop request */ ereport(FATAL, @@ -7310,6 +7317,32 @@ xlog_redo(XLogRecPtr lsn, XLogRecord *record) { /* nothing to do here */ } + else if (info == XLOG_BACKUP_END) + { + XLogRecPtr startpoint; + memcpy(&startpoint, XLogRecGetData(record), sizeof(startpoint)); + + if (XLByteEQ(ControlFile->backupStartPoint, startpoint)) + { + /* + * We have reached the end of base backup, the point where + * pg_stop_backup() was done. The data on disk is now consistent. + * Reset backupStartPoint, and update minRecoveryPoint to make + * sure we don't allow starting up at an earlier point even if + * recovery is stopped and restarted soon after this. + */ + elog(DEBUG1, "end of backup reached"); + + LWLockAcquire(ControlFileLock, LW_EXCLUSIVE); + + if (XLByteLT(ControlFile->minRecoveryPoint, lsn)) + ControlFile->minRecoveryPoint = lsn; + MemSet(&ControlFile->backupStartPoint, 0, sizeof(XLogRecPtr)); + UpdateControlFile(); + + LWLockRelease(ControlFileLock); + } + } } void @@ -7351,6 +7384,14 @@ xlog_desc(StringInfo buf, uint8 xl_info, char *rec) { appendStringInfo(buf, "xlog switch"); } + else if (info == XLOG_BACKUP_END) + { + XLogRecPtr startpoint; + + memcpy(&startpoint, rec, sizeof(XLogRecPtr)); + appendStringInfo(buf, "backup end: %X/%X", + startpoint.xlogid, startpoint.xrecoff); + } else appendStringInfo(buf, "UNKNOWN"); } @@ -7686,10 +7727,14 @@ pg_start_backup_callback(int code, Datum arg) /* * pg_stop_backup: finish taking an on-line backup dump * - * We remove the backup label file created by pg_start_backup, and instead - * create a backup history file in pg_xlog (whence it will immediately be - * archived). The backup history file contains the same info found in - * the label file, plus the backup-end time and WAL location. + * We write an end-of-backup WAL record, and remove the backup label file + * created by pg_start_backup, creating a backup history file in pg_xlog + * instead (whence it will immediately be archived). The backup history file + * contains the same info found in the label file, plus the backup-end time + * and WAL location. Before 8.5, the backup-end time was read from the backup + * history file at the beginning of archive recovery, but we now use the WAL + * record for that and the file is for informational and debug purposes only. + * * Note: different from CancelBackup which just cancels online backup mode. */ Datum @@ -7697,6 +7742,7 @@ pg_stop_backup(PG_FUNCTION_ARGS) { XLogRecPtr startpoint; XLogRecPtr stoppoint; + XLogRecData rdata; pg_time_t stamp_time; char strfbuf[128]; char histfilepath[MAXPGPATH]; @@ -7738,22 +7784,6 @@ pg_stop_backup(PG_FUNCTION_ARGS) LWLockRelease(WALInsertLock); /* - * Force a switch to a new xlog segment file, so that the backup is valid - * as soon as archiver moves out the current segment file. We'll report - * the end address of the XLOG SWITCH record as the backup stopping point. - */ - stoppoint = RequestXLogSwitch(); - - XLByteToSeg(stoppoint, _logId, _logSeg); - XLogFileName(stopxlogfilename, ThisTimeLineID, _logId, _logSeg); - - /* Use the log timezone here, not the session timezone */ - stamp_time = (pg_time_t) time(NULL); - pg_strftime(strfbuf, sizeof(strfbuf), - "%Y-%m-%d %H:%M:%S %Z", - pg_localtime(&stamp_time, log_timezone)); - - /* * Open the existing label file */ lfp = AllocateFile(BACKUP_LABEL_FILE, "r"); @@ -7781,6 +7811,30 @@ pg_stop_backup(PG_FUNCTION_ARGS) errmsg("invalid data in file \"%s\"", BACKUP_LABEL_FILE))); /* + * Write the backup-end xlog record + */ + rdata.data = (char *) (&startpoint); + rdata.len = sizeof(startpoint); + rdata.buffer = InvalidBuffer; + rdata.next = NULL; + stoppoint = XLogInsert(RM_XLOG_ID, XLOG_BACKUP_END, &rdata); + + /* + * Force a switch to a new xlog segment file, so that the backup is valid + * as soon as archiver moves out the current segment file. + */ + RequestXLogSwitch(); + + XLByteToSeg(stoppoint, _logId, _logSeg); + XLogFileName(stopxlogfilename, ThisTimeLineID, _logId, _logSeg); + + /* Use the log timezone here, not the session timezone */ + stamp_time = (pg_time_t) time(NULL); + pg_strftime(strfbuf, sizeof(strfbuf), + "%Y-%m-%d %H:%M:%S %Z", + pg_localtime(&stamp_time, log_timezone)); + + /* * Write the backup history file */ XLByteToSeg(startpoint, _logId, _logSeg); @@ -8086,33 +8140,18 @@ pg_xlogfile_name(PG_FUNCTION_ARGS) * later than the start of the dump, and so if we rely on it as the start * point, we will fail to restore a consistent database state. * - * We also attempt to retrieve the corresponding backup history file. - * If successful, set *minRecoveryLoc to constrain valid PITR stopping - * points. - * * Returns TRUE if a backup_label was found (and fills the checkpoint * location into *checkPointLoc); returns FALSE if not. */ static bool -read_backup_label(XLogRecPtr *checkPointLoc, XLogRecPtr *minRecoveryLoc) +read_backup_label(XLogRecPtr *checkPointLoc) { XLogRecPtr startpoint; - XLogRecPtr stoppoint; - char histfilename[MAXFNAMELEN]; - char histfilepath[MAXPGPATH]; char startxlogfilename[MAXFNAMELEN]; - char stopxlogfilename[MAXFNAMELEN]; TimeLineID tli; - uint32 _logId; - uint32 _logSeg; FILE *lfp; - FILE *fp; char ch; - /* Default is to not constrain recovery stop point */ - minRecoveryLoc->xlogid = 0; - minRecoveryLoc->xrecoff = 0; - /* * See if label file is present */ @@ -8150,45 +8189,6 @@ read_backup_label(XLogRecPtr *checkPointLoc, XLogRecPtr *minRecoveryLoc) errmsg("could not read file \"%s\": %m", BACKUP_LABEL_FILE))); - /* - * Try to retrieve the backup history file (no error if we can't) - */ - XLByteToSeg(startpoint, _logId, _logSeg); - BackupHistoryFileName(histfilename, tli, _logId, _logSeg, - startpoint.xrecoff % XLogSegSize); - - if (InArchiveRecovery) - RestoreArchivedFile(histfilepath, histfilename, "RECOVERYHISTORY", 0); - else - BackupHistoryFilePath(histfilepath, tli, _logId, _logSeg, - startpoint.xrecoff % XLogSegSize); - - fp = AllocateFile(histfilepath, "r"); - if (fp) - { - /* - * Parse history file to identify stop point. - */ - if (fscanf(fp, "START WAL LOCATION: %X/%X (file %24s)%c", - &startpoint.xlogid, &startpoint.xrecoff, startxlogfilename, - &ch) != 4 || ch != '\n') - ereport(FATAL, - (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), - errmsg("invalid data in file \"%s\"", histfilename))); - if (fscanf(fp, "STOP WAL LOCATION: %X/%X (file %24s)%c", - &stoppoint.xlogid, &stoppoint.xrecoff, stopxlogfilename, - &ch) != 4 || ch != '\n') - ereport(FATAL, - (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), - errmsg("invalid data in file \"%s\"", histfilename))); - *minRecoveryLoc = stoppoint; - if (ferror(fp) || FreeFile(fp)) - ereport(FATAL, - (errcode_for_file_access(), - errmsg("could not read file \"%s\": %m", - histfilepath))); - } - return true; } diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h index 2435f27..31cc537 100644 --- a/src/include/catalog/pg_control.h +++ b/src/include/catalog/pg_control.h @@ -62,6 +62,7 @@ typedef struct CheckPoint #define XLOG_NOOP 0x20 #define XLOG_NEXTOID 0x30 #define XLOG_SWITCH 0x40 +#define XLOG_BACKUP_END 0x50 /* System status indicator */ @@ -117,7 +118,27 @@ typedef struct ControlFileData CheckPoint checkPointCopy; /* copy of last check point record */ - XLogRecPtr minRecoveryPoint; /* must replay xlog to here */ + /* + * These two values determine the minimum point we must recover up to + * before starting up: + * + * minRecoveryPoint is updated to the latest replayed LSN whenever we + * flush a data change during archive recovery. That guards against + * starting archive recovery, aborting it, and restarting with an earlier + * stop location. If we've already flushed to disk data changes from WAL + * record, we mustn't start up until we reach X again. Should be zero + * when we're not doing archive recovery. + * + * backupStartPoint is the redo pointer of the backup start checkpoint, + * if we are recovering from an online backup and haven't reached the end + * of backup yet. It is reset to 0 when the end of backup is reached, and + * we mustn't start up before that. A boolean would suffice otherwise, but + * we use the redo location as a cross-check when we see an end-of-backup + * record, to make sure the end-of-backup record is the counterpart of + * the base backup actually used. + */ + XLogRecPtr minRecoveryPoint; + XLogRecPtr backupStartPoint; /* * This data is used to check for hardware-architecture compatibility of
pgsql-hackers by date: