pg_resetxlog: add option to set oldest xid & use by pg_upgrade

Add pg_resetxlog -u option to set the oldest xid in pg_control.
Previously -x set this value be -2 billion less than the -x value.
However, this causes the server to immediately scan all relation's
relfrozenxid so it can advance pg_control's oldest xid to be inside the
autovacuum_freeze_max_age range, which is inefficient and might disrupt
diagnostic recovery.  pg_upgrade will use this option to better create
the new cluster to match the old cluster.

Reported-by: Jason Harvey, Floris Van Nee

Discussion: https://postgr.es/m/20190615183759.GB239428@rfd.leadboat.com, 87da83168c644fd9aae38f546cc70295@opammb0562.comp.optiver.com

Author: Bertrand Drouvot

Backpatch-through: 9.6

doc/src/sgml/ref/pg_resetwal.sgml

@@ -297,6 +297,26 @@ PostgreSQL documentation
     </listitem>
    </varlistentry>
 
+   <varlistentry>
+    <term><option>-u <replaceable class="parameter">xid</replaceable></option></term>
+    <term><option>--oldest-transaction-id=<replaceable class="parameter">xid</replaceable></option></term>
+    <listitem>
+     <para>
+      Manually set the oldest unfrozen transaction ID.
+     </para>
+
+     <para>
+      A safe value can be determined by looking for the numerically smallest
+      file name in the directory <filename>pg_xact</filename> under the data directory
+      and then multiplying by 1048576 (0x100000).  Note that the file names are in
+      hexadecimal.  It is usually easiest to specify the option value in
+      hexadecimal too. For example, if <filename>0007</filename> is the smallest entry
+      in <filename>pg_xact</filename>, <literal>-u 0x700000</literal> will work (five
+      trailing zeroes provide the proper multiplier).
+     </para>
+    </listitem>
+   </varlistentry>
+
    <varlistentry>
     <term><option>-x <replaceable class="parameter">xid</replaceable></option></term>
     <term><option>--next-transaction-id=<replaceable class="parameter">xid</replaceable></option></term>

src/bin/pg_resetwal/pg_resetwal.c

@@ -64,6 +64,7 @@ static XLogSegNo newXlogSegNo;	/* new XLOG segment # */
 static bool guessed = false;	/* T if we had to guess at any values */
 static const char *progname;
 static uint32 set_xid_epoch = (uint32) -1;
+static TransactionId set_oldest_xid = 0;
 static TransactionId set_xid = 0;
 static TransactionId set_oldest_commit_ts_xid = 0;
 static TransactionId set_newest_commit_ts_xid = 0;
@@ -101,6 +102,7 @@ main(int argc, char *argv[])
 		{"dry-run", no_argument, NULL, 'n'},
 		{"next-oid", required_argument, NULL, 'o'},
 		{"multixact-offset", required_argument, NULL, 'O'},
+		{"oldest-transaction-id", required_argument, NULL, 'u'},
 		{"next-transaction-id", required_argument, NULL, 'x'},
 		{"wal-segsize", required_argument, NULL, 1},
 		{NULL, 0, NULL, 0}
@@ -135,7 +137,7 @@ main(int argc, char *argv[])
 	}
 
 
-	while ((c = getopt_long(argc, argv, "c:D:e:fl:m:no:O:x:", long_options, NULL)) != -1)
+	while ((c = getopt_long(argc, argv, "c:D:e:fl:m:no:O:u:x:", long_options, NULL)) != -1)
 	{
 		switch (c)
 		{
@@ -168,6 +170,21 @@ main(int argc, char *argv[])
 				}
 				break;
 
+			case 'u':
+				set_oldest_xid = strtoul(optarg, &endptr, 0);
+				if (endptr == optarg || *endptr != '\0')
+				{
+					pg_log_error("invalid argument for option %s", "-u");
+					fprintf(stderr, _("Try \"%s --help\" for more information.\n"), progname);
+					exit(1);
+				}
+				if (!TransactionIdIsNormal(set_oldest_xid))
+				{
+					pg_log_error("oldest transaction ID (-u) must be greater or equal to %u", FirstNormalTransactionId);
+					exit(1);
+				}
+				break;
+
 			case 'x':
 				set_xid = strtoul(optarg, &endptr, 0);
 				if (endptr == optarg || *endptr != '\0')
@@ -176,9 +193,9 @@ main(int argc, char *argv[])
 					fprintf(stderr, _("Try \"%s --help\" for more information.\n"), progname);
 					exit(1);
 				}
-				if (set_xid == 0)
+				if (!TransactionIdIsNormal(set_xid))
 				{
-					pg_log_error("transaction ID (-x) must not be 0");
+					pg_log_error("transaction ID (-x) must be greater or equal to %u", FirstNormalTransactionId);
 					exit(1);
 				}
 				break;
@@ -428,25 +445,17 @@ main(int argc, char *argv[])
 			FullTransactionIdFromEpochAndXid(set_xid_epoch,
 											 XidFromFullTransactionId(ControlFile.checkPointCopy.nextXid));
 
-	if (set_xid != 0)
+	if (set_oldest_xid != 0)
 	{
+		ControlFile.checkPointCopy.oldestXid = set_oldest_xid;
+		ControlFile.checkPointCopy.oldestXidDB = InvalidOid;
+	}
+
+	if (set_xid != 0)
 		ControlFile.checkPointCopy.nextXid =
 			FullTransactionIdFromEpochAndXid(EpochFromFullTransactionId(ControlFile.checkPointCopy.nextXid),
 											 set_xid);
 
-		/*
-		 * For the moment, just set oldestXid to a value that will force
-		 * immediate autovacuum-for-wraparound.  It's not clear whether adding
-		 * user control of this is useful, so let's just do something that's
-		 * reasonably safe.  The magic constant here corresponds to the
-		 * maximum allowed value of autovacuum_freeze_max_age.
-		 */
-		ControlFile.checkPointCopy.oldestXid = set_xid - 2000000000;
-		if (ControlFile.checkPointCopy.oldestXid < FirstNormalTransactionId)
-			ControlFile.checkPointCopy.oldestXid += FirstNormalTransactionId;
-		ControlFile.checkPointCopy.oldestXidDB = InvalidOid;
-	}
-
 	if (set_oldest_commit_ts_xid != 0)
 		ControlFile.checkPointCopy.oldestCommitTsXid = set_oldest_commit_ts_xid;
 	if (set_newest_commit_ts_xid != 0)
@@ -1209,20 +1218,21 @@ usage(void)
 	printf(_("Usage:\n  %s [OPTION]... DATADIR\n\n"), progname);
 	printf(_("Options:\n"));
 	printf(_("  -c, --commit-timestamp-ids=XID,XID\n"
-			 "                                 set oldest and newest transactions bearing\n"
+			 "                                   set oldest and newest transactions bearing\n"
-			 "                                 commit timestamp (zero means no change)\n"));
+			 "                                   commit timestamp (zero means no change)\n"));
-	printf(_(" [-D, --pgdata=]DATADIR          data directory\n"));
+	printf(_(" [-D, --pgdata=]DATADIR            data directory\n"));
-	printf(_("  -e, --epoch=XIDEPOCH           set next transaction ID epoch\n"));
+	printf(_("  -e, --epoch=XIDEPOCH             set next transaction ID epoch\n"));
-	printf(_("  -f, --force                    force update to be done\n"));
+	printf(_("  -f, --force                      force update to be done\n"));
-	printf(_("  -l, --next-wal-file=WALFILE    set minimum starting location for new WAL\n"));
+	printf(_("  -l, --next-wal-file=WALFILE      set minimum starting location for new WAL\n"));
-	printf(_("  -m, --multixact-ids=MXID,MXID  set next and oldest multitransaction ID\n"));
+	printf(_("  -m, --multixact-ids=MXID,MXID    set next and oldest multitransaction ID\n"));
-	printf(_("  -n, --dry-run                  no update, just show what would be done\n"));
+	printf(_("  -n, --dry-run                    no update, just show what would be done\n"));
-	printf(_("  -o, --next-oid=OID             set next OID\n"));
+	printf(_("  -o, --next-oid=OID               set next OID\n"));
-	printf(_("  -O, --multixact-offset=OFFSET  set next multitransaction offset\n"));
+	printf(_("  -O, --multixact-offset=OFFSET    set next multitransaction offset\n"));
-	printf(_("  -V, --version                  output version information, then exit\n"));
+	printf(_("  -u, --oldest-transaction-id=XID  set oldest transaction ID\n"));
-	printf(_("  -x, --next-transaction-id=XID  set next transaction ID\n"));
+	printf(_("  -V, --version                    output version information, then exit\n"));
-	printf(_("      --wal-segsize=SIZE         size of WAL segments, in megabytes\n"));
+	printf(_("  -x, --next-transaction-id=XID    set next transaction ID\n"));
-	printf(_("  -?, --help                     show this help, then exit\n"));
+	printf(_("      --wal-segsize=SIZE           size of WAL segments, in megabytes\n"));
+	printf(_("  -?, --help                       show this help, then exit\n"));
 	printf(_("\nReport bugs to <%s>.\n"), PACKAGE_BUGREPORT);
 	printf(_("%s home page: <%s>\n"), PACKAGE_NAME, PACKAGE_URL);
 }

src/bin/pg_upgrade/controldata.c

@@ -44,6 +44,7 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 	bool		got_oid = false;
 	bool		got_multi = false;
 	bool		got_oldestmulti = false;
+	bool		got_oldestxid = false;
 	bool		got_mxoff = false;
 	bool		got_nextxlogfile = false;
 	bool		got_float8_pass_by_value = false;
@@ -312,6 +313,17 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 			cluster->controldata.chkpnt_nxtmulti = str2uint(p);
 			got_multi = true;
 		}
+		else if ((p = strstr(bufin, "Latest checkpoint's oldestXID:")) != NULL)
+		{
+			p = strchr(p, ':');
+
+			if (p == NULL || strlen(p) <= 1)
+				pg_fatal("%d: controldata retrieval problem\n", __LINE__);
+
+			p++;				/* remove ':' char */
+			cluster->controldata.chkpnt_oldstxid = str2uint(p);
+			got_oldestxid = true;
+		}
 		else if ((p = strstr(bufin, "Latest checkpoint's oldestMultiXid:")) != NULL)
 		{
 			p = strchr(p, ':');
@@ -544,7 +556,7 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 
 	/* verify that we got all the mandatory pg_control data */
 	if (!got_xid || !got_oid ||
-		!got_multi ||
+		!got_multi || !got_oldestxid ||
 		(!got_oldestmulti &&
 		 cluster->controldata.cat_ver >= MULTIXACT_FORMATCHANGE_CAT_VER) ||
 		!got_mxoff || (!live_check && !got_nextxlogfile) ||
@@ -575,6 +587,9 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 			cluster->controldata.cat_ver >= MULTIXACT_FORMATCHANGE_CAT_VER)
 			pg_log(PG_REPORT, "  latest checkpoint oldest MultiXactId\n");
 
+		if (!got_oldestxid)
+			pg_log(PG_REPORT, "  latest checkpoint oldestXID\n");
+
 		if (!got_mxoff)
 			pg_log(PG_REPORT, "  latest checkpoint next MultiXactOffset\n");
 

src/bin/pg_upgrade/pg_upgrade.c

@@ -467,6 +467,13 @@ copy_xact_xlog_xid(void)
 					  GET_MAJOR_VERSION(new_cluster.major_version) <= 906 ?
 					  "pg_clog" : "pg_xact");
 
+	prep_status("Setting oldest XID for new cluster");
+	exec_prog(UTILITY_LOG_FILE, NULL, true, true,
+			  "\"%s/pg_resetwal\" -f -u %u \"%s\"",
+			  new_cluster.bindir, old_cluster.controldata.chkpnt_oldstxid,
+			  new_cluster.pgdata);
+	check_ok();
+
 	/* set the next transaction id and epoch of the new cluster */
 	prep_status("Setting next transaction ID and epoch for new cluster");
 	exec_prog(UTILITY_LOG_FILE, NULL, true, true,

src/bin/pg_upgrade/pg_upgrade.h

@@ -207,6 +207,7 @@ typedef struct
 	uint32		chkpnt_nxtmulti;
 	uint32		chkpnt_nxtmxoff;
 	uint32		chkpnt_oldstMulti;
+	uint32		chkpnt_oldstxid;
 	uint32		align;
 	uint32		blocksz;
 	uint32		largesz;