2015-03-23 18:47:52 +01:00
|
|
|
/*-------------------------------------------------------------------------
|
|
|
|
*
|
|
|
|
* pg_rewind.c
|
2015-12-01 16:56:44 +01:00
|
|
|
* Synchronizes a PostgreSQL data directory to a new timeline
|
2015-03-23 18:47:52 +01:00
|
|
|
*
|
2018-01-03 05:30:12 +01:00
|
|
|
* Portions Copyright (c) 1996-2018, PostgreSQL Global Development Group
|
2015-03-23 18:47:52 +01:00
|
|
|
*
|
|
|
|
*-------------------------------------------------------------------------
|
|
|
|
*/
|
|
|
|
#include "postgres_fe.h"
|
|
|
|
|
|
|
|
#include <sys/stat.h>
|
|
|
|
#include <fcntl.h>
|
|
|
|
#include <time.h>
|
|
|
|
#include <unistd.h>
|
|
|
|
|
|
|
|
#include "pg_rewind.h"
|
|
|
|
#include "fetch.h"
|
|
|
|
#include "file_ops.h"
|
|
|
|
#include "filemap.h"
|
|
|
|
#include "logging.h"
|
|
|
|
|
|
|
|
#include "access/timeline.h"
|
|
|
|
#include "access/xlog_internal.h"
|
|
|
|
#include "catalog/catversion.h"
|
|
|
|
#include "catalog/pg_control.h"
|
2018-04-07 23:45:39 +02:00
|
|
|
#include "common/file_perm.h"
|
2018-07-10 01:39:27 +02:00
|
|
|
#include "common/file_utils.h"
|
2015-04-07 22:04:25 +02:00
|
|
|
#include "common/restricted_token.h"
|
2015-03-23 18:47:52 +01:00
|
|
|
#include "getopt_long.h"
|
|
|
|
#include "storage/bufpage.h"
|
|
|
|
|
|
|
|
static void usage(const char *progname);
|
|
|
|
|
|
|
|
static void createBackupLabel(XLogRecPtr startpoint, TimeLineID starttli,
|
|
|
|
XLogRecPtr checkpointloc);
|
|
|
|
|
|
|
|
static void digestControlFile(ControlFileData *ControlFile, char *source,
|
|
|
|
size_t size);
|
|
|
|
static void updateControlFile(ControlFileData *ControlFile);
|
2016-03-27 23:46:25 +02:00
|
|
|
static void syncTargetDirectory(const char *argv0);
|
2015-03-23 18:47:52 +01:00
|
|
|
static void sanityChecks(void);
|
2015-12-01 16:56:44 +01:00
|
|
|
static void findCommonAncestorTimeline(XLogRecPtr *recptr, int *tliIndex);
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
static ControlFileData ControlFile_target;
|
|
|
|
static ControlFileData ControlFile_source;
|
|
|
|
|
|
|
|
const char *progname;
|
Make WAL segment size configurable at initdb time.
For performance reasons a larger segment size than the default 16MB
can be useful. A larger segment size has two main benefits: Firstly,
in setups using archiving, it makes it easier to write scripts that
can keep up with higher amounts of WAL, secondly, the WAL has to be
written and synced to disk less frequently.
But at the same time large segment size are disadvantageous for
smaller databases. So far the segment size had to be configured at
compile time, often making it unrealistic to choose one fitting to a
particularly load. Therefore change it to a initdb time setting.
This includes a breaking changes to the xlogreader.h API, which now
requires the current segment size to be configured. For that and
similar reasons a number of binaries had to be taught how to recognize
the current segment size.
Author: Beena Emerson, editorialized by Andres Freund
Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael
Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja
Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
|
|
|
int WalSegSz;
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
/* Configuration options */
|
|
|
|
char *datadir_target = NULL;
|
|
|
|
char *datadir_source = NULL;
|
|
|
|
char *connstr_source = NULL;
|
|
|
|
|
|
|
|
bool debug = false;
|
|
|
|
bool showprogress = false;
|
|
|
|
bool dry_run = false;
|
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
/* Target history */
|
|
|
|
TimeLineHistoryEntry *targetHistory;
|
2016-06-10 00:02:36 +02:00
|
|
|
int targetNentries;
|
2015-12-01 16:56:44 +01:00
|
|
|
|
2015-03-23 18:47:52 +01:00
|
|
|
static void
|
|
|
|
usage(const char *progname)
|
|
|
|
{
|
2015-06-23 02:40:01 +02:00
|
|
|
printf(_("%s resynchronizes a PostgreSQL cluster with another copy of the cluster.\n\n"), progname);
|
2015-03-23 18:47:52 +01:00
|
|
|
printf(_("Usage:\n %s [OPTION]...\n\n"), progname);
|
|
|
|
printf(_("Options:\n"));
|
2015-06-23 02:40:01 +02:00
|
|
|
printf(_(" -D, --target-pgdata=DIRECTORY existing data directory to modify\n"));
|
2015-09-16 06:37:39 +02:00
|
|
|
printf(_(" --source-pgdata=DIRECTORY source data directory to synchronize with\n"));
|
|
|
|
printf(_(" --source-server=CONNSTR source server to synchronize with\n"));
|
2015-06-23 02:40:01 +02:00
|
|
|
printf(_(" -n, --dry-run stop before modifying anything\n"));
|
|
|
|
printf(_(" -P, --progress write progress messages\n"));
|
|
|
|
printf(_(" --debug write a lot of debug messages\n"));
|
|
|
|
printf(_(" -V, --version output version information, then exit\n"));
|
|
|
|
printf(_(" -?, --help show this help, then exit\n"));
|
|
|
|
printf(_("\nReport bugs to <pgsql-bugs@postgresql.org>.\n"));
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
int
|
|
|
|
main(int argc, char **argv)
|
|
|
|
{
|
|
|
|
static struct option long_options[] = {
|
|
|
|
{"help", no_argument, NULL, '?'},
|
|
|
|
{"target-pgdata", required_argument, NULL, 'D'},
|
|
|
|
{"source-pgdata", required_argument, NULL, 1},
|
|
|
|
{"source-server", required_argument, NULL, 2},
|
|
|
|
{"version", no_argument, NULL, 'V'},
|
|
|
|
{"dry-run", no_argument, NULL, 'n'},
|
|
|
|
{"progress", no_argument, NULL, 'P'},
|
|
|
|
{"debug", no_argument, NULL, 3},
|
|
|
|
{NULL, 0, NULL, 0}
|
|
|
|
};
|
|
|
|
int option_index;
|
|
|
|
int c;
|
|
|
|
XLogRecPtr divergerec;
|
2015-12-01 16:56:44 +01:00
|
|
|
int lastcommontliIndex;
|
2015-03-23 18:47:52 +01:00
|
|
|
XLogRecPtr chkptrec;
|
|
|
|
TimeLineID chkpttli;
|
|
|
|
XLogRecPtr chkptredo;
|
|
|
|
size_t size;
|
|
|
|
char *buffer;
|
|
|
|
bool rewind_needed;
|
|
|
|
XLogRecPtr endrec;
|
|
|
|
TimeLineID endtli;
|
|
|
|
ControlFileData ControlFile_new;
|
|
|
|
|
2015-04-07 22:04:25 +02:00
|
|
|
set_pglocale_pgservice(argv[0], PG_TEXTDOMAIN("pg_rewind"));
|
2015-03-23 18:47:52 +01:00
|
|
|
progname = get_progname(argv[0]);
|
|
|
|
|
|
|
|
/* Process command-line arguments */
|
|
|
|
if (argc > 1)
|
|
|
|
{
|
|
|
|
if (strcmp(argv[1], "--help") == 0 || strcmp(argv[1], "-?") == 0)
|
|
|
|
{
|
|
|
|
usage(progname);
|
|
|
|
exit(0);
|
|
|
|
}
|
|
|
|
if (strcmp(argv[1], "--version") == 0 || strcmp(argv[1], "-V") == 0)
|
|
|
|
{
|
|
|
|
puts("pg_rewind (PostgreSQL) " PG_VERSION);
|
|
|
|
exit(0);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-06-11 15:31:18 +02:00
|
|
|
while ((c = getopt_long(argc, argv, "D:nP", long_options, &option_index)) != -1)
|
2015-03-23 18:47:52 +01:00
|
|
|
{
|
|
|
|
switch (c)
|
|
|
|
{
|
|
|
|
case '?':
|
|
|
|
fprintf(stderr, _("Try \"%s --help\" for more information.\n"), progname);
|
|
|
|
exit(1);
|
|
|
|
|
|
|
|
case 'P':
|
|
|
|
showprogress = true;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case 'n':
|
|
|
|
dry_run = true;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case 3:
|
|
|
|
debug = true;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case 'D': /* -D or --target-pgdata */
|
|
|
|
datadir_target = pg_strdup(optarg);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case 1: /* --source-pgdata */
|
|
|
|
datadir_source = pg_strdup(optarg);
|
|
|
|
break;
|
|
|
|
case 2: /* --source-server */
|
|
|
|
connstr_source = pg_strdup(optarg);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (datadir_source == NULL && connstr_source == NULL)
|
|
|
|
{
|
2015-06-23 02:40:01 +02:00
|
|
|
fprintf(stderr, _("%s: no source specified (--source-pgdata or --source-server)\n"), progname);
|
2015-04-07 22:28:28 +02:00
|
|
|
fprintf(stderr, _("Try \"%s --help\" for more information.\n"), progname);
|
2015-03-23 18:47:52 +01:00
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
|
2016-10-07 13:35:17 +02:00
|
|
|
if (datadir_source != NULL && connstr_source != NULL)
|
|
|
|
{
|
|
|
|
fprintf(stderr, _("%s: only one of --source-pgdata or --source-server can be specified\n"), progname);
|
|
|
|
fprintf(stderr, _("Try \"%s --help\" for more information.\n"), progname);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
|
2015-03-23 18:47:52 +01:00
|
|
|
if (datadir_target == NULL)
|
|
|
|
{
|
2015-06-23 02:40:01 +02:00
|
|
|
fprintf(stderr, _("%s: no target data directory specified (--target-pgdata)\n"), progname);
|
2015-03-23 18:47:52 +01:00
|
|
|
fprintf(stderr, _("Try \"%s --help\" for more information.\n"), progname);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
|
2015-06-23 02:40:01 +02:00
|
|
|
if (optind < argc)
|
2015-03-23 18:47:52 +01:00
|
|
|
{
|
2015-06-23 02:40:01 +02:00
|
|
|
fprintf(stderr, _("%s: too many command-line arguments (first is \"%s\")\n"),
|
|
|
|
progname, argv[optind]);
|
2015-03-23 18:47:52 +01:00
|
|
|
fprintf(stderr, _("Try \"%s --help\" for more information.\n"), progname);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
|
2015-04-07 22:04:25 +02:00
|
|
|
/*
|
|
|
|
* Don't allow pg_rewind to be run as root, to avoid overwriting the
|
|
|
|
* ownership of files in the data directory. We need only check for root
|
|
|
|
* -- any other user won't have sufficient permissions to modify files in
|
|
|
|
* the data directory.
|
|
|
|
*/
|
|
|
|
#ifndef WIN32
|
|
|
|
if (geteuid() == 0)
|
2015-06-23 02:40:01 +02:00
|
|
|
{
|
|
|
|
fprintf(stderr, _("cannot be executed by \"root\"\n"));
|
|
|
|
fprintf(stderr, _("You must run %s as the PostgreSQL superuser.\n"),
|
|
|
|
progname);
|
2018-04-09 21:33:33 +02:00
|
|
|
exit(1);
|
2015-06-23 02:40:01 +02:00
|
|
|
}
|
2015-04-07 22:04:25 +02:00
|
|
|
#endif
|
|
|
|
|
|
|
|
get_restricted_token(progname);
|
|
|
|
|
2018-05-23 16:59:55 +02:00
|
|
|
/* Set mask based on PGDATA permissions */
|
|
|
|
if (!GetDataDirectoryCreatePerm(datadir_target))
|
|
|
|
{
|
|
|
|
fprintf(stderr, _("%s: could not read permissions of directory \"%s\": %s\n"),
|
|
|
|
progname, datadir_target, strerror(errno));
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
|
|
|
|
umask(pg_mode_mask);
|
|
|
|
|
2015-03-23 18:47:52 +01:00
|
|
|
/* Connect to remote server */
|
|
|
|
if (connstr_source)
|
|
|
|
libpqConnect(connstr_source);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Ok, we have all the options and we're ready to start. Read in all the
|
|
|
|
* information we need from both clusters.
|
|
|
|
*/
|
|
|
|
buffer = slurpFile(datadir_target, "global/pg_control", &size);
|
|
|
|
digestControlFile(&ControlFile_target, buffer, size);
|
|
|
|
pg_free(buffer);
|
|
|
|
|
|
|
|
buffer = fetchFile("global/pg_control", &size);
|
|
|
|
digestControlFile(&ControlFile_source, buffer, size);
|
|
|
|
pg_free(buffer);
|
|
|
|
|
|
|
|
sanityChecks();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If both clusters are already on the same timeline, there's nothing to
|
|
|
|
* do.
|
|
|
|
*/
|
|
|
|
if (ControlFile_target.checkPointCopy.ThisTimeLineID == ControlFile_source.checkPointCopy.ThisTimeLineID)
|
|
|
|
{
|
2015-12-04 04:21:16 +01:00
|
|
|
printf(_("source and target cluster are on the same timeline\n"));
|
|
|
|
rewind_needed = false;
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2015-12-04 04:21:16 +01:00
|
|
|
findCommonAncestorTimeline(&divergerec, &lastcommontliIndex);
|
2017-05-12 19:51:27 +02:00
|
|
|
printf(_("servers diverged at WAL location %X/%X on timeline %u\n"),
|
2015-12-04 04:21:16 +01:00
|
|
|
(uint32) (divergerec >> 32), (uint32) divergerec,
|
|
|
|
targetHistory[lastcommontliIndex].tli);
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
/*
|
2016-06-10 00:02:36 +02:00
|
|
|
* Check for the possibility that the target is in fact a direct
|
|
|
|
* ancestor of the source. In that case, there is no divergent history
|
|
|
|
* in the target that needs rewinding.
|
2015-03-23 18:47:52 +01:00
|
|
|
*/
|
2015-12-04 04:21:16 +01:00
|
|
|
if (ControlFile_target.checkPoint >= divergerec)
|
|
|
|
{
|
2015-03-23 18:47:52 +01:00
|
|
|
rewind_needed = true;
|
2015-12-04 04:21:16 +01:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
XLogRecPtr chkptendrec;
|
|
|
|
|
|
|
|
/* Read the checkpoint record on the target to see where it ends. */
|
|
|
|
chkptendrec = readOneRecord(datadir_target,
|
|
|
|
ControlFile_target.checkPoint,
|
|
|
|
targetNentries - 1);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the histories diverged exactly at the end of the shutdown
|
2016-06-10 00:02:36 +02:00
|
|
|
* checkpoint record on the target, there are no WAL records in
|
|
|
|
* the target that don't belong in the source's history, and no
|
|
|
|
* rewind is needed.
|
2015-12-04 04:21:16 +01:00
|
|
|
*/
|
|
|
|
if (chkptendrec == divergerec)
|
|
|
|
rewind_needed = false;
|
|
|
|
else
|
|
|
|
rewind_needed = true;
|
|
|
|
}
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!rewind_needed)
|
|
|
|
{
|
2015-10-02 03:42:00 +02:00
|
|
|
printf(_("no rewind required\n"));
|
2015-03-23 18:47:52 +01:00
|
|
|
exit(0);
|
|
|
|
}
|
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
findLastCheckpoint(datadir_target, divergerec,
|
|
|
|
lastcommontliIndex,
|
2015-03-23 18:47:52 +01:00
|
|
|
&chkptrec, &chkpttli, &chkptredo);
|
2015-10-02 03:42:00 +02:00
|
|
|
printf(_("rewinding from last common checkpoint at %X/%X on timeline %u\n"),
|
2015-03-23 18:47:52 +01:00
|
|
|
(uint32) (chkptrec >> 32), (uint32) chkptrec,
|
|
|
|
chkpttli);
|
|
|
|
|
|
|
|
/*
|
2015-04-15 21:52:00 +02:00
|
|
|
* Build the filemap, by comparing the source and target data directories.
|
2015-03-23 18:47:52 +01:00
|
|
|
*/
|
2015-03-30 02:02:14 +02:00
|
|
|
filemap_create();
|
2015-03-23 18:47:52 +01:00
|
|
|
pg_log(PG_PROGRESS, "reading source file list\n");
|
2015-04-15 21:52:00 +02:00
|
|
|
fetchSourceFileList();
|
2015-03-23 18:47:52 +01:00
|
|
|
pg_log(PG_PROGRESS, "reading target file list\n");
|
2015-04-15 21:52:00 +02:00
|
|
|
traverse_datadir(datadir_target, &process_target_file);
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Read the target WAL from last checkpoint before the point of fork, to
|
|
|
|
* extract all the pages that were modified on the target cluster after
|
|
|
|
* the fork. We can stop reading after reaching the final shutdown record.
|
|
|
|
* XXX: If we supported rewinding a server that was not shut down cleanly,
|
|
|
|
* we would need to replay until the end of WAL here.
|
|
|
|
*/
|
|
|
|
pg_log(PG_PROGRESS, "reading WAL in target\n");
|
2015-12-01 16:56:44 +01:00
|
|
|
extractPageMap(datadir_target, chkptrec, lastcommontliIndex,
|
2015-03-23 18:47:52 +01:00
|
|
|
ControlFile_target.checkPoint);
|
|
|
|
filemap_finalize();
|
|
|
|
|
|
|
|
if (showprogress)
|
|
|
|
calculate_totals();
|
|
|
|
|
|
|
|
/* this is too verbose even for verbose mode */
|
|
|
|
if (debug)
|
|
|
|
print_filemap();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Ok, we're ready to start copying things over.
|
|
|
|
*/
|
|
|
|
if (showprogress)
|
|
|
|
{
|
2015-06-23 02:40:01 +02:00
|
|
|
pg_log(PG_PROGRESS, "need to copy %lu MB (total source directory size is %lu MB)\n",
|
2015-03-23 18:47:52 +01:00
|
|
|
(unsigned long) (filemap->fetch_size / (1024 * 1024)),
|
|
|
|
(unsigned long) (filemap->total_size / (1024 * 1024)));
|
|
|
|
|
|
|
|
fetch_size = filemap->fetch_size;
|
|
|
|
fetch_done = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is the point of no return. Once we start copying things, we have
|
|
|
|
* modified the target directory and there is no turning back!
|
|
|
|
*/
|
|
|
|
|
|
|
|
executeFileMap();
|
|
|
|
|
|
|
|
progress_report(true);
|
|
|
|
|
|
|
|
pg_log(PG_PROGRESS, "\ncreating backup label and updating control file\n");
|
|
|
|
createBackupLabel(chkptredo, chkpttli, chkptrec);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update control file of target. Make it ready to perform archive
|
|
|
|
* recovery when restarting.
|
|
|
|
*
|
|
|
|
* minRecoveryPoint is set to the current WAL insert location in the
|
|
|
|
* source server. Like in an online backup, it's important that we recover
|
|
|
|
* all the WAL that was generated while we copied the files over.
|
|
|
|
*/
|
|
|
|
memcpy(&ControlFile_new, &ControlFile_source, sizeof(ControlFileData));
|
|
|
|
|
|
|
|
if (connstr_source)
|
|
|
|
{
|
|
|
|
endrec = libpqGetCurrentXlogInsertLocation();
|
|
|
|
endtli = ControlFile_source.checkPointCopy.ThisTimeLineID;
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
endrec = ControlFile_source.checkPoint;
|
|
|
|
endtli = ControlFile_source.checkPointCopy.ThisTimeLineID;
|
|
|
|
}
|
|
|
|
ControlFile_new.minRecoveryPoint = endrec;
|
|
|
|
ControlFile_new.minRecoveryPointTLI = endtli;
|
|
|
|
ControlFile_new.state = DB_IN_ARCHIVE_RECOVERY;
|
|
|
|
updateControlFile(&ControlFile_new);
|
|
|
|
|
2016-03-27 23:46:25 +02:00
|
|
|
pg_log(PG_PROGRESS, "syncing target data directory\n");
|
|
|
|
syncTargetDirectory(argv[0]);
|
|
|
|
|
2015-03-23 18:47:52 +01:00
|
|
|
printf(_("Done!\n"));
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
sanityChecks(void)
|
|
|
|
{
|
|
|
|
/* TODO Check that there's no backup_label in either cluster */
|
|
|
|
|
|
|
|
/* Check system_id match */
|
|
|
|
if (ControlFile_target.system_identifier != ControlFile_source.system_identifier)
|
|
|
|
pg_fatal("source and target clusters are from different systems\n");
|
|
|
|
|
|
|
|
/* check version */
|
|
|
|
if (ControlFile_target.pg_control_version != PG_CONTROL_VERSION ||
|
|
|
|
ControlFile_source.pg_control_version != PG_CONTROL_VERSION ||
|
|
|
|
ControlFile_target.catalog_version_no != CATALOG_VERSION_NO ||
|
|
|
|
ControlFile_source.catalog_version_no != CATALOG_VERSION_NO)
|
|
|
|
{
|
|
|
|
pg_fatal("clusters are not compatible with this version of pg_rewind\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Target cluster need to use checksums or hint bit wal-logging, this to
|
|
|
|
* prevent from data corruption that could occur because of hint bits.
|
|
|
|
*/
|
|
|
|
if (ControlFile_target.data_checksum_version != PG_DATA_CHECKSUM_VERSION &&
|
|
|
|
!ControlFile_target.wal_log_hints)
|
|
|
|
{
|
2015-06-11 15:31:18 +02:00
|
|
|
pg_fatal("target server needs to use either data checksums or \"wal_log_hints = on\"\n");
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Target cluster better not be running. This doesn't guard against
|
|
|
|
* someone starting the cluster concurrently. Also, this is probably more
|
2015-12-01 16:56:44 +01:00
|
|
|
* strict than necessary; it's OK if the target node was not shut down
|
|
|
|
* cleanly, as long as it isn't running at the moment.
|
2015-03-23 18:47:52 +01:00
|
|
|
*/
|
2015-12-01 16:56:44 +01:00
|
|
|
if (ControlFile_target.state != DB_SHUTDOWNED &&
|
|
|
|
ControlFile_target.state != DB_SHUTDOWNED_IN_RECOVERY)
|
2015-03-23 18:47:52 +01:00
|
|
|
pg_fatal("target server must be shut down cleanly\n");
|
|
|
|
|
|
|
|
/*
|
|
|
|
* When the source is a data directory, also require that the source
|
|
|
|
* server is shut down. There isn't any very strong reason for this
|
|
|
|
* limitation, but better safe than sorry.
|
|
|
|
*/
|
2015-12-01 16:56:44 +01:00
|
|
|
if (datadir_source &&
|
|
|
|
ControlFile_source.state != DB_SHUTDOWNED &&
|
|
|
|
ControlFile_source.state != DB_SHUTDOWNED_IN_RECOVERY)
|
2015-03-23 18:47:52 +01:00
|
|
|
pg_fatal("source data directory must be shut down cleanly\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2017-05-12 19:51:27 +02:00
|
|
|
* Find minimum from two WAL locations assuming InvalidXLogRecPtr means
|
2015-12-01 16:56:44 +01:00
|
|
|
* infinity as src/include/access/timeline.h states. This routine should
|
2017-05-12 19:51:27 +02:00
|
|
|
* be used only when comparing WAL locations related to history files.
|
2015-03-23 18:47:52 +01:00
|
|
|
*/
|
2015-12-01 16:56:44 +01:00
|
|
|
static XLogRecPtr
|
|
|
|
MinXLogRecPtr(XLogRecPtr a, XLogRecPtr b)
|
2015-03-23 18:47:52 +01:00
|
|
|
{
|
2015-12-01 16:56:44 +01:00
|
|
|
if (XLogRecPtrIsInvalid(a))
|
|
|
|
return b;
|
|
|
|
else if (XLogRecPtrIsInvalid(b))
|
|
|
|
return a;
|
|
|
|
else
|
|
|
|
return Min(a, b);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Retrieve timeline history for given control file which should behold
|
|
|
|
* either source or target.
|
|
|
|
*/
|
|
|
|
static TimeLineHistoryEntry *
|
|
|
|
getTimelineHistory(ControlFileData *controlFile, int *nentries)
|
|
|
|
{
|
2016-06-10 00:02:36 +02:00
|
|
|
TimeLineHistoryEntry *history;
|
|
|
|
TimeLineID tli;
|
2015-03-23 18:47:52 +01:00
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
tli = controlFile->checkPointCopy.ThisTimeLineID;
|
2015-03-23 18:47:52 +01:00
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
/*
|
2016-06-10 00:02:36 +02:00
|
|
|
* Timeline 1 does not have a history file, so there is no need to check
|
|
|
|
* and fake an entry with infinite start and end positions.
|
2015-12-01 16:56:44 +01:00
|
|
|
*/
|
|
|
|
if (tli == 1)
|
2015-03-23 18:47:52 +01:00
|
|
|
{
|
2015-12-01 16:56:44 +01:00
|
|
|
history = (TimeLineHistoryEntry *) pg_malloc(sizeof(TimeLineHistoryEntry));
|
|
|
|
history->tli = tli;
|
|
|
|
history->begin = history->end = InvalidXLogRecPtr;
|
|
|
|
*nentries = 1;
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
char path[MAXPGPATH];
|
|
|
|
char *histfile;
|
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
TLHistoryFilePath(path, tli);
|
|
|
|
|
|
|
|
/* Get history file from appropriate source */
|
|
|
|
if (controlFile == &ControlFile_source)
|
|
|
|
histfile = fetchFile(path, NULL);
|
|
|
|
else if (controlFile == &ControlFile_target)
|
|
|
|
histfile = slurpFile(datadir_target, path, NULL);
|
|
|
|
else
|
2016-06-07 20:18:08 +02:00
|
|
|
pg_fatal("invalid control file");
|
2015-03-23 18:47:52 +01:00
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
history = rewind_parseTimeLineHistory(histfile, tli, nentries);
|
2015-03-23 18:47:52 +01:00
|
|
|
pg_free(histfile);
|
|
|
|
}
|
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
if (debug)
|
2015-03-23 18:47:52 +01:00
|
|
|
{
|
2016-06-10 00:02:36 +02:00
|
|
|
int i;
|
2015-12-01 16:56:44 +01:00
|
|
|
|
|
|
|
if (controlFile == &ControlFile_source)
|
2016-03-28 19:33:00 +02:00
|
|
|
pg_log(PG_DEBUG, "Source timeline history:\n");
|
2015-12-01 16:56:44 +01:00
|
|
|
else if (controlFile == &ControlFile_target)
|
2016-03-28 19:33:00 +02:00
|
|
|
pg_log(PG_DEBUG, "Target timeline history:\n");
|
2015-12-01 16:56:44 +01:00
|
|
|
else
|
|
|
|
Assert(false);
|
2015-03-23 18:47:52 +01:00
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
/*
|
|
|
|
* Print the target timeline history.
|
|
|
|
*/
|
|
|
|
for (i = 0; i < targetNentries; i++)
|
2015-03-23 18:47:52 +01:00
|
|
|
{
|
2015-12-01 16:56:44 +01:00
|
|
|
TimeLineHistoryEntry *entry;
|
2015-03-23 18:47:52 +01:00
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
entry = &history[i];
|
2016-03-28 19:33:00 +02:00
|
|
|
pg_log(PG_DEBUG,
|
|
|
|
/* translator: %d is a timeline number, others are LSN positions */
|
|
|
|
"%d: %X/%X - %X/%X\n", entry->tli,
|
|
|
|
(uint32) (entry->begin >> 32), (uint32) (entry->begin),
|
|
|
|
(uint32) (entry->end >> 32), (uint32) (entry->end));
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-12-01 16:56:44 +01:00
|
|
|
return history;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Determine the TLI of the last common timeline in the timeline history of the
|
|
|
|
* two clusters. targetHistory is filled with target timeline history and
|
|
|
|
* targetNentries is number of items in targetHistory. *tliIndex is set to the
|
|
|
|
* index of last common timeline in targetHistory array, and *recptr is set to
|
|
|
|
* the position where the timeline history diverged (ie. the first WAL record
|
|
|
|
* that's not the same in both clusters).
|
|
|
|
*
|
|
|
|
* Control files of both clusters must be read into ControlFile_target/source
|
|
|
|
* before calling this routine.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
findCommonAncestorTimeline(XLogRecPtr *recptr, int *tliIndex)
|
|
|
|
{
|
|
|
|
TimeLineHistoryEntry *sourceHistory;
|
|
|
|
int sourceNentries;
|
2016-06-10 00:02:36 +02:00
|
|
|
int i,
|
|
|
|
n;
|
2015-12-01 16:56:44 +01:00
|
|
|
|
|
|
|
/* Retrieve timelines for both source and target */
|
|
|
|
sourceHistory = getTimelineHistory(&ControlFile_source, &sourceNentries);
|
|
|
|
targetHistory = getTimelineHistory(&ControlFile_target, &targetNentries);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Trace the history forward, until we hit the timeline diverge. It may
|
|
|
|
* still be possible that the source and target nodes used the same
|
|
|
|
* timeline number in their history but with different start position
|
|
|
|
* depending on the history files that each node has fetched in previous
|
|
|
|
* recovery processes. Hence check the start position of the new timeline
|
|
|
|
* as well and move down by one extra timeline entry if they do not match.
|
|
|
|
*/
|
|
|
|
n = Min(sourceNentries, targetNentries);
|
|
|
|
for (i = 0; i < n; i++)
|
|
|
|
{
|
|
|
|
if (sourceHistory[i].tli != targetHistory[i].tli ||
|
|
|
|
sourceHistory[i].begin != targetHistory[i].begin)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (i > 0)
|
|
|
|
{
|
|
|
|
i--;
|
|
|
|
*recptr = MinXLogRecPtr(sourceHistory[i].end, targetHistory[i].end);
|
|
|
|
*tliIndex = i;
|
|
|
|
|
|
|
|
pg_free(sourceHistory);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
pg_fatal("could not find common ancestor of the source and target cluster's timelines\n");
|
|
|
|
}
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Create a backup_label file that forces recovery to begin at the last common
|
|
|
|
* checkpoint.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
createBackupLabel(XLogRecPtr startpoint, TimeLineID starttli, XLogRecPtr checkpointloc)
|
|
|
|
{
|
|
|
|
XLogSegNo startsegno;
|
|
|
|
time_t stamp_time;
|
|
|
|
char strfbuf[128];
|
|
|
|
char xlogfilename[MAXFNAMELEN];
|
|
|
|
struct tm *tmp;
|
|
|
|
char buf[1000];
|
|
|
|
int len;
|
|
|
|
|
Make WAL segment size configurable at initdb time.
For performance reasons a larger segment size than the default 16MB
can be useful. A larger segment size has two main benefits: Firstly,
in setups using archiving, it makes it easier to write scripts that
can keep up with higher amounts of WAL, secondly, the WAL has to be
written and synced to disk less frequently.
But at the same time large segment size are disadvantageous for
smaller databases. So far the segment size had to be configured at
compile time, often making it unrealistic to choose one fitting to a
particularly load. Therefore change it to a initdb time setting.
This includes a breaking changes to the xlogreader.h API, which now
requires the current segment size to be configured. For that and
similar reasons a number of binaries had to be taught how to recognize
the current segment size.
Author: Beena Emerson, editorialized by Andres Freund
Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael
Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja
Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
|
|
|
XLByteToSeg(startpoint, startsegno, WalSegSz);
|
|
|
|
XLogFileName(xlogfilename, starttli, startsegno, WalSegSz);
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Construct backup label file
|
|
|
|
*/
|
|
|
|
stamp_time = time(NULL);
|
|
|
|
tmp = localtime(&stamp_time);
|
|
|
|
strftime(strfbuf, sizeof(strfbuf), "%Y-%m-%d %H:%M:%S %Z", tmp);
|
|
|
|
|
|
|
|
len = snprintf(buf, sizeof(buf),
|
|
|
|
"START WAL LOCATION: %X/%X (file %s)\n"
|
|
|
|
"CHECKPOINT LOCATION: %X/%X\n"
|
|
|
|
"BACKUP METHOD: pg_rewind\n"
|
|
|
|
"BACKUP FROM: standby\n"
|
|
|
|
"START TIME: %s\n",
|
2015-05-24 03:35:49 +02:00
|
|
|
/* omit LABEL: line */
|
Phase 3 of pgindent updates.
Don't move parenthesized lines to the left, even if that means they
flow past the right margin.
By default, BSD indent lines up statement continuation lines that are
within parentheses so that they start just to the right of the preceding
left parenthesis. However, traditionally, if that resulted in the
continuation line extending to the right of the desired right margin,
then indent would push it left just far enough to not overrun the margin,
if it could do so without making the continuation line start to the left of
the current statement indent. That makes for a weird mix of indentations
unless one has been completely rigid about never violating the 80-column
limit.
This behavior has been pretty universally panned by Postgres developers.
Hence, disable it with indent's new -lpl switch, so that parenthesized
lines are always lined up with the preceding left paren.
This patch is much less interesting than the first round of indent
changes, but also bulkier, so I thought it best to separate the effects.
Discussion: https://postgr.es/m/E1dAmxK-0006EE-1r@gemulon.postgresql.org
Discussion: https://postgr.es/m/30527.1495162840@sss.pgh.pa.us
2017-06-21 21:35:54 +02:00
|
|
|
(uint32) (startpoint >> 32), (uint32) startpoint, xlogfilename,
|
2015-03-23 18:47:52 +01:00
|
|
|
(uint32) (checkpointloc >> 32), (uint32) checkpointloc,
|
|
|
|
strfbuf);
|
|
|
|
if (len >= sizeof(buf))
|
2015-05-24 03:35:49 +02:00
|
|
|
pg_fatal("backup label buffer too small\n"); /* shouldn't happen */
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
/* TODO: move old file out of the way, if any. */
|
Phase 2 of pgindent updates.
Change pg_bsd_indent to follow upstream rules for placement of comments
to the right of code, and remove pgindent hack that caused comments
following #endif to not obey the general rule.
Commit e3860ffa4dd0dad0dd9eea4be9cc1412373a8c89 wasn't actually using
the published version of pg_bsd_indent, but a hacked-up version that
tried to minimize the amount of movement of comments to the right of
code. The situation of interest is where such a comment has to be
moved to the right of its default placement at column 33 because there's
code there. BSD indent has always moved right in units of tab stops
in such cases --- but in the previous incarnation, indent was working
in 8-space tab stops, while now it knows we use 4-space tabs. So the
net result is that in about half the cases, such comments are placed
one tab stop left of before. This is better all around: it leaves
more room on the line for comment text, and it means that in such
cases the comment uniformly starts at the next 4-space tab stop after
the code, rather than sometimes one and sometimes two tabs after.
Also, ensure that comments following #endif are indented the same
as comments following other preprocessor commands such as #else.
That inconsistency turns out to have been self-inflicted damage
from a poorly-thought-through post-indent "fixup" in pgindent.
This patch is much less interesting than the first round of indent
changes, but also bulkier, so I thought it best to separate the effects.
Discussion: https://postgr.es/m/E1dAmxK-0006EE-1r@gemulon.postgresql.org
Discussion: https://postgr.es/m/30527.1495162840@sss.pgh.pa.us
2017-06-21 21:18:54 +02:00
|
|
|
open_target_file("backup_label", true); /* BACKUP_LABEL_FILE */
|
2015-03-23 18:47:52 +01:00
|
|
|
write_target_range(buf, 0, len);
|
2016-03-27 22:48:31 +02:00
|
|
|
close_target_file();
|
2015-03-23 18:47:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check CRC of control file
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
checkControlFile(ControlFileData *ControlFile)
|
|
|
|
{
|
2015-04-14 16:03:42 +02:00
|
|
|
pg_crc32c crc;
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
/* Calculate CRC */
|
|
|
|
INIT_CRC32C(crc);
|
|
|
|
COMP_CRC32C(crc, (char *) ControlFile, offsetof(ControlFileData, crc));
|
|
|
|
FIN_CRC32C(crc);
|
|
|
|
|
|
|
|
/* And simply compare it */
|
|
|
|
if (!EQ_CRC32C(crc, ControlFile->crc))
|
|
|
|
pg_fatal("unexpected control file CRC\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Verify control file contents in the buffer src, and copy it to *ControlFile.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
digestControlFile(ControlFileData *ControlFile, char *src, size_t size)
|
|
|
|
{
|
2017-07-19 22:16:57 +02:00
|
|
|
if (size != PG_CONTROL_FILE_SIZE)
|
2015-03-23 18:47:52 +01:00
|
|
|
pg_fatal("unexpected control file size %d, expected %d\n",
|
2017-07-19 22:16:57 +02:00
|
|
|
(int) size, PG_CONTROL_FILE_SIZE);
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
memcpy(ControlFile, src, sizeof(ControlFileData));
|
|
|
|
|
Make WAL segment size configurable at initdb time.
For performance reasons a larger segment size than the default 16MB
can be useful. A larger segment size has two main benefits: Firstly,
in setups using archiving, it makes it easier to write scripts that
can keep up with higher amounts of WAL, secondly, the WAL has to be
written and synced to disk less frequently.
But at the same time large segment size are disadvantageous for
smaller databases. So far the segment size had to be configured at
compile time, often making it unrealistic to choose one fitting to a
particularly load. Therefore change it to a initdb time setting.
This includes a breaking changes to the xlogreader.h API, which now
requires the current segment size to be configured. For that and
similar reasons a number of binaries had to be taught how to recognize
the current segment size.
Author: Beena Emerson, editorialized by Andres Freund
Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael
Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja
Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
|
|
|
/* set and validate WalSegSz */
|
|
|
|
WalSegSz = ControlFile->xlog_seg_size;
|
|
|
|
|
|
|
|
if (!IsValidWalSegSize(WalSegSz))
|
2018-05-18 05:05:27 +02:00
|
|
|
pg_fatal(ngettext("WAL segment size must be a power of two between 1 MB and 1 GB, but the control file specifies %d byte\n",
|
|
|
|
"WAL segment size must be a power of two between 1 MB and 1 GB, but the control file specifies %d bytes\n",
|
|
|
|
WalSegSz),
|
Make WAL segment size configurable at initdb time.
For performance reasons a larger segment size than the default 16MB
can be useful. A larger segment size has two main benefits: Firstly,
in setups using archiving, it makes it easier to write scripts that
can keep up with higher amounts of WAL, secondly, the WAL has to be
written and synced to disk less frequently.
But at the same time large segment size are disadvantageous for
smaller databases. So far the segment size had to be configured at
compile time, often making it unrealistic to choose one fitting to a
particularly load. Therefore change it to a initdb time setting.
This includes a breaking changes to the xlogreader.h API, which now
requires the current segment size to be configured. For that and
similar reasons a number of binaries had to be taught how to recognize
the current segment size.
Author: Beena Emerson, editorialized by Andres Freund
Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael
Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja
Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
|
|
|
WalSegSz);
|
|
|
|
|
2015-03-23 18:47:52 +01:00
|
|
|
/* Additional checks on control file */
|
|
|
|
checkControlFile(ControlFile);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the target's control file.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
updateControlFile(ControlFileData *ControlFile)
|
|
|
|
{
|
2017-07-19 22:16:57 +02:00
|
|
|
char buffer[PG_CONTROL_FILE_SIZE];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For good luck, apply the same static assertions as in backend's
|
|
|
|
* WriteControlFile().
|
|
|
|
*/
|
|
|
|
StaticAssertStmt(sizeof(ControlFileData) <= PG_CONTROL_MAX_SAFE_SIZE,
|
|
|
|
"pg_control is too large for atomic disk writes");
|
|
|
|
StaticAssertStmt(sizeof(ControlFileData) <= PG_CONTROL_FILE_SIZE,
|
|
|
|
"sizeof(ControlFileData) exceeds PG_CONTROL_FILE_SIZE");
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
/* Recalculate CRC of control file */
|
|
|
|
INIT_CRC32C(ControlFile->crc);
|
|
|
|
COMP_CRC32C(ControlFile->crc,
|
|
|
|
(char *) ControlFile,
|
|
|
|
offsetof(ControlFileData, crc));
|
|
|
|
FIN_CRC32C(ControlFile->crc);
|
|
|
|
|
|
|
|
/*
|
2017-07-19 22:16:57 +02:00
|
|
|
* Write out PG_CONTROL_FILE_SIZE bytes into pg_control by zero-padding
|
|
|
|
* the excess over sizeof(ControlFileData), to avoid premature EOF related
|
2015-03-23 18:47:52 +01:00
|
|
|
* errors when reading it.
|
|
|
|
*/
|
2017-07-19 22:16:57 +02:00
|
|
|
memset(buffer, 0, PG_CONTROL_FILE_SIZE);
|
2015-03-23 18:47:52 +01:00
|
|
|
memcpy(buffer, ControlFile, sizeof(ControlFileData));
|
|
|
|
|
|
|
|
open_target_file("global/pg_control", false);
|
|
|
|
|
2017-07-19 22:16:57 +02:00
|
|
|
write_target_range(buffer, 0, PG_CONTROL_FILE_SIZE);
|
2015-03-23 18:47:52 +01:00
|
|
|
|
|
|
|
close_target_file();
|
|
|
|
}
|
2016-03-27 23:46:25 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Sync target data directory to ensure that modifications are safely on disk.
|
|
|
|
*
|
|
|
|
* We do this once, for the whole data directory, for performance reasons. At
|
|
|
|
* the end of pg_rewind's run, the kernel is likely to already have flushed
|
2018-07-10 01:39:27 +02:00
|
|
|
* most dirty buffers to disk. Additionally fsync_pgdata uses a two-pass
|
|
|
|
* approach (only initiating writeback in the first pass), which often reduces
|
|
|
|
* the overall amount of IO noticeably.
|
2016-03-27 23:46:25 +02:00
|
|
|
*/
|
|
|
|
static void
|
|
|
|
syncTargetDirectory(const char *argv0)
|
|
|
|
{
|
|
|
|
if (dry_run)
|
|
|
|
return;
|
|
|
|
|
2018-07-10 01:39:27 +02:00
|
|
|
fsync_pgdata(datadir_target, progname, PG_VERSION_NUM);
|
2016-03-27 23:46:25 +02:00
|
|
|
}
|