Skip to content

Commit 5d2b45e

Browse files
MasaoFujiiMasaoFujii
committed
Add GUC to control the time to wait before retrieving WAL after failed attempt.
Previously when the standby server failed to retrieve WAL files from any sources (i.e., streaming replication, local pg_xlog directory or WAL archive), it always waited for five seconds (hard-coded) before the next attempt. For example, this is problematic in warm-standby because restore_command can fail every five seconds even while new WAL file is expected to be unavailable for a long time and flood the log files with its error messages. This commit adds new parameter, wal_retrieve_retry_interval, to control that wait time. Alexey Vasiliev and Michael Paquier, reviewed by Andres Freund and me.
1 parent 2a3f6e3 commit 5d2b45e

File tree

5 files changed

+57
-13
lines changed

5 files changed

+57
-13
lines changed

‎doc/src/sgml/config.sgml

Lines changed: 18 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2985,6 +2985,24 @@ include_dir 'conf.d'
29852985
</listitem>
29862986
</varlistentry>
29872987

2988+
<varlistentry id="guc-wal-retrieve-retry-interval" xreflabel="wal_retrieve_retry_interval">
2989+
<term><varname>wal_retrieve_retry_interval</varname> (<type>integer</type>)
2990+
<indexterm>
2991+
<primary><varname>wal_retrieve_retry_interval</> configuration parameter</primary>
2992+
</indexterm>
2993+
</term>
2994+
<listitem>
2995+
<para>
2996+
Specify how long the standby server should wait when WAL data is not
2997+
available from any sources (streaming replication,
2998+
local <filename>pg_xlog</> or WAL archive) before retrying to
2999+
retrieve WAL data. This parameter can only be set in the
3000+
<filename>postgresql.conf</> file or on the server command line.
3001+
The default value is 5 seconds. Units are milliseconds if not specified.
3002+
</para>
3003+
</listitem>
3004+
</varlistentry>
3005+
29883006
</variablelist>
29893007
</sect2>
29903008
</sect1>

‎src/backend/access/transam/xlog.c

Lines changed: 24 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -93,6 +93,7 @@ int sync_method = DEFAULT_SYNC_METHOD;
9393
int wal_level = WAL_LEVEL_MINIMAL;
9494
int CommitDelay = 0; /* precommit delay in microseconds */
9595
int CommitSiblings = 5; /* # concurrent xacts needed to sleep */
96+
int wal_retrieve_retry_interval = 5000;
9697

9798
#ifdef WAL_DEBUG
9899
bool XLOG_DEBUG = false;
@@ -10340,8 +10341,8 @@ static bool
1034010341
WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
1034110342
bool fetching_ckpt, XLogRecPtr tliRecPtr)
1034210343
{
10343-
static pg_time_t last_fail_time = 0;
10344-
pg_time_t now;
10344+
static TimestampTz last_fail_time = 0;
10345+
TimestampTz now;
1034510346

1034610347
/*-------
1034710348
* Standby mode is implemented by a state machine:
@@ -10351,7 +10352,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
1035110352
* 2. Check trigger file
1035210353
* 3. Read from primary server via walreceiver (XLOG_FROM_STREAM)
1035310354
* 4. Rescan timelines
10354-
* 5. Sleep 5 seconds, and loop back to 1.
10355+
* 5. Sleep wal_retrieve_retry_interval milliseconds, and loop back to 1.
1035510356
*
1035610357
* Failure to read from the current source advances the state machine to
1035710358
* the next state.
@@ -10490,14 +10491,25 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
1049010491
* machine, so we've exhausted all the options for
1049110492
* obtaining the requested WAL. We're going to loop back
1049210493
* and retry from the archive, but if it hasn't been long
10493-
* since last attempt, sleep 5 seconds to avoid
10494-
* busy-waiting.
10494+
* since last attempt, sleep wal_retrieve_retry_interval
10495+
* milliseconds to avoid busy-waiting.
1049510496
*/
10496-
now = (pg_time_t) time(NULL);
10497-
if ((now - last_fail_time) < 5)
10497+
now = GetCurrentTimestamp();
10498+
if (!TimestampDifferenceExceeds(last_fail_time, now,
10499+
wal_retrieve_retry_interval))
1049810500
{
10499-
pg_usleep(1000000L * (5 - (now - last_fail_time)));
10500-
now = (pg_time_t) time(NULL);
10501+
long secs, wait_time;
10502+
int usecs;
10503+
10504+
TimestampDifference(last_fail_time, now, &secs, &usecs);
10505+
wait_time = wal_retrieve_retry_interval -
10506+
(secs * 1000 + usecs / 1000);
10507+
10508+
WaitLatch(&XLogCtl->recoveryWakeupLatch,
10509+
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
10510+
wait_time);
10511+
ResetLatch(&XLogCtl->recoveryWakeupLatch);
10512+
now = GetCurrentTimestamp();
1050110513
}
1050210514
last_fail_time = now;
1050310515
currentSource = XLOG_FROM_ARCHIVE;
@@ -10653,12 +10665,11 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
1065310665
}
1065410666

1065510667
/*
10656-
* Wait for more WAL to arrive. Time out after 5 seconds,
10657-
* like when polling the archive, to react to a trigger
10658-
* file promptly.
10668+
* Wait for more WAL to arrive. Time out after 5 seconds
10669+
* to react to a trigger file promptly.
1065910670
*/
1066010671
WaitLatch(&XLogCtl->recoveryWakeupLatch,
10661-
WL_LATCH_SET | WL_TIMEOUT,
10672+
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
1066210673
5000L);
1066310674
ResetLatch(&XLogCtl->recoveryWakeupLatch);
1066410675
break;

‎src/backend/utils/misc/guc.c

Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2363,6 +2363,18 @@ static struct config_int ConfigureNamesInt[] =
23632363
NULL, NULL, NULL
23642364
},
23652365

2366+
{
2367+
{"wal_retrieve_retry_interval", PGC_SIGHUP, REPLICATION_STANDBY,
2368+
gettext_noop("Sets the time to wait before retrying to retrieve WAL"
2369+
"after a failed attempt."),
2370+
NULL,
2371+
GUC_UNIT_MS
2372+
},
2373+
&wal_retrieve_retry_interval,
2374+
5000, 1, INT_MAX,
2375+
NULL, NULL, NULL
2376+
},
2377+
23662378
{
23672379
{"wal_segment_size", PGC_INTERNAL, PRESET_OPTIONS,
23682380
gettext_noop("Shows the number of pages per write ahead log segment."),

‎src/backend/utils/misc/postgresql.conf.sample

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -260,6 +260,8 @@
260260
#wal_receiver_timeout = 60s # time that receiver waits for
261261
# communication from master
262262
# in milliseconds; 0 disables
263+
#wal_retrieve_retry_interval = 5s # time to wait before retrying to
264+
# retrieve WAL after a failed attempt
263265

264266

265267
#------------------------------------------------------------------------------

‎src/include/access/xlog.h

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -93,6 +93,7 @@ extern int CheckPointSegments;
9393
extern int wal_keep_segments;
9494
extern int XLOGbuffers;
9595
extern int XLogArchiveTimeout;
96+
extern int wal_retrieve_retry_interval;
9697
extern bool XLogArchiveMode;
9798
extern char *XLogArchiveCommand;
9899
extern bool EnableHotStandby;

0 commit comments

Comments
 (0)