Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit5cbbe70

Browse files
committed
Flush the IO statistics of active WAL senders more frequently
WAL senders do not flush their statistics until they exit, limiting themonitoring possible for live processes. This is penalizing when WALsenders are running for a long time, like in streaming or logicalreplication setups, because it is not possible to know the amount of IOthey generate while running.This commit makes WAL senders more aggressive with their statisticsflush, using an internal of 1 second, with the flush timing calculatedbased on the existing GetCurrentTimestamp() done before the sleeps doneto wait for some activity. Note that the sleep done for logical andphysical WAL senders happens in two different code paths, so the statsflushes need to happen in these two places.One test is added for the physical WAL sender case, and one for thelogical WAL sender case. This can be done in a stable fashion byrelying on the WAL generated by the TAP tests in combination with astats reset while a server is running, but only on HEAD as WAL data hasbeen added to pg_stat_io ina051e71.This issue exists sincea9c70b4 and the introduction of pg_stat_io,so backpatch down to v16.Author: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>Reviewed-by: vignesh C <vignesh21@gmail.com>Reviewed-by: Xuneng Zhou <xunengzhou@gmail.com>Discussion:https://postgr.es/m/Z73IsKBceoVd4t55@ip-10-97-1-34.eu-west-3.compute.internalBackpatch-through: 16
1 parent4c1d853 commit5cbbe70

File tree

1 file changed

+32
-2
lines changed

1 file changed

+32
-2
lines changed

‎src/backend/replication/walsender.c

Lines changed: 32 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -90,10 +90,14 @@
9090
#include"utils/guc.h"
9191
#include"utils/memutils.h"
9292
#include"utils/pg_lsn.h"
93+
#include"utils/pgstat_internal.h"
9394
#include"utils/ps_status.h"
9495
#include"utils/timeout.h"
9596
#include"utils/timestamp.h"
9697

98+
/* Minimum interval used by walsender for stats flushes, in ms */
99+
#defineWALSENDER_STATS_FLUSH_INTERVAL 1000
100+
97101
/*
98102
* Maximum data payload in a WAL data message. Must be >= XLOG_BLCKSZ.
99103
*
@@ -1820,6 +1824,7 @@ WalSndWaitForWal(XLogRecPtr loc)
18201824
intwakeEvents;
18211825
uint32wait_event=0;
18221826
staticXLogRecPtrRecentFlushPtr=InvalidXLogRecPtr;
1827+
TimestampTzlast_flush=0;
18231828

18241829
/*
18251830
* Fast path to avoid acquiring the spinlock in case we already know we
@@ -1840,6 +1845,7 @@ WalSndWaitForWal(XLogRecPtr loc)
18401845
{
18411846
boolwait_for_standby_at_stop= false;
18421847
longsleeptime;
1848+
TimestampTznow;
18431849

18441850
/* Clear any already-pending wakeups */
18451851
ResetLatch(MyLatch);
@@ -1950,7 +1956,8 @@ WalSndWaitForWal(XLogRecPtr loc)
19501956
* new WAL to be generated. (But if we have nothing to send, we don't
19511957
* want to wake on socket-writable.)
19521958
*/
1953-
sleeptime=WalSndComputeSleeptime(GetCurrentTimestamp());
1959+
now=GetCurrentTimestamp();
1960+
sleeptime=WalSndComputeSleeptime(now);
19541961

19551962
wakeEvents=WL_SOCKET_READABLE;
19561963

@@ -1959,6 +1966,14 @@ WalSndWaitForWal(XLogRecPtr loc)
19591966

19601967
Assert(wait_event!=0);
19611968

1969+
/* Report IO statistics, if needed */
1970+
if (TimestampDifferenceExceeds(last_flush,now,
1971+
WALSENDER_STATS_FLUSH_INTERVAL))
1972+
{
1973+
pgstat_flush_io(false);
1974+
last_flush=now;
1975+
}
1976+
19621977
WalSndWait(wakeEvents,sleeptime,wait_event);
19631978
}
19641979

@@ -2766,6 +2781,8 @@ WalSndCheckTimeOut(void)
27662781
staticvoid
27672782
WalSndLoop(WalSndSendDataCallbacksend_data)
27682783
{
2784+
TimestampTzlast_flush=0;
2785+
27692786
/*
27702787
* Initialize the last reply timestamp. That enables timeout processing
27712788
* from hereon.
@@ -2860,13 +2877,17 @@ WalSndLoop(WalSndSendDataCallback send_data)
28602877
* WalSndWaitForWal() handle any other blocking; idle receivers need
28612878
* its additional actions. For physical replication, also block if
28622879
* caught up; its send_data does not block.
2880+
*
2881+
* The IO statistics are reported in WalSndWaitForWal() for the
2882+
* logical WAL senders.
28632883
*/
28642884
if ((WalSndCaughtUp&&send_data!=XLogSendLogical&&
28652885
!streamingDoneSending)||
28662886
pq_is_send_pending())
28672887
{
28682888
longsleeptime;
28692889
intwakeEvents;
2890+
TimestampTznow;
28702891

28712892
if (!streamingDoneReceiving)
28722893
wakeEvents=WL_SOCKET_READABLE;
@@ -2877,11 +2898,20 @@ WalSndLoop(WalSndSendDataCallback send_data)
28772898
* Use fresh timestamp, not last_processing, to reduce the chance
28782899
* of reaching wal_sender_timeout before sending a keepalive.
28792900
*/
2880-
sleeptime=WalSndComputeSleeptime(GetCurrentTimestamp());
2901+
now=GetCurrentTimestamp();
2902+
sleeptime=WalSndComputeSleeptime(now);
28812903

28822904
if (pq_is_send_pending())
28832905
wakeEvents |=WL_SOCKET_WRITEABLE;
28842906

2907+
/* Report IO statistics, if needed */
2908+
if (TimestampDifferenceExceeds(last_flush,now,
2909+
WALSENDER_STATS_FLUSH_INTERVAL))
2910+
{
2911+
pgstat_flush_io(false);
2912+
last_flush=now;
2913+
}
2914+
28852915
/* Sleep until something happens or we time out */
28862916
WalSndWait(wakeEvents,sleeptime,WAIT_EVENT_WAL_SENDER_MAIN);
28872917
}

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp