Hi,
I use Postgres 17.7 and I have configured incremental backup
wal_summary_keep_time=10d
show wal_summary_keep_time=on
On the standby, this message is written every 10 seconds by the
walsummarizer process.
2026-02-17 11:40:22.548 CET [544677]: [297-1]
user=,db=,client=,application= ERROR: could not read WAL from timeline 6
at 698/460000A0: invalid record length at 698/460000A0: expected at least
24, got 0
2026-02-17 11:40:32.559 CET [544677]: [298-1]
user=,db=,client=,application= ERROR: could not read WAL from timeline 6
at 698/460000A0: invalid record length at 698/460000A0: expected at least
24, got 0
postgres 544677 544615 0 10:50 ? 00:00:00 postgres: dev_ge:
walsummarizer
Last summary files are:
-rw-------. 1 postgres postgres 1808 Feb 5 18:00
00000007000006998F2205F0000006998F2FCAA8.summary
-rw-------. 1 postgres postgres 1158 Feb 5 18:30
00000007000006998F2FCAA8000006998F328C48.summary
-rw-------. 1 postgres postgres 448 Feb 5 19:00
00000007000006998F328C48000006998F3419C0.summary
-rw-------. 1 postgres postgres 566 Feb 5 19:30
00000007000006998F3419C0000006998F35D2C8.summary
-rw-------. 1 postgres postgres 616 Feb 5 20:00
00000007000006998F35D2C8000006998F379C60.summary
-rw-------. 1 postgres postgres 1108 Feb 5 20:30
00000007000006998F379C60000006998F3A88C8.summary
-rw-------. 1 postgres postgres 330 Feb 5 20:43
00000007000006998F3A88C80000069990000028.summary
-rw-------. 1 postgres postgres 32 Feb 5 20:43
00000007000006999000002800000699900000A0.summary
-rw-------. 1 postgres postgres 32 Feb 5 20:43
0000000800000699900000A000000699900000D8.summary
summary file of the timeline 6
-rw-------. 1 postgres postgres 780 Jan 8 20:10
0000000600000698452AB9680000069846000028.summary
-rw-------. 1 postgres postgres 32 Jan 8 20:10
00000006000006984600002800000698460000A0.summary
pg_waldump 000000060000069800000046
rmgr: XLOG len (rec/tot): 114/ 114, tx: 0, lsn:
698/46000028, prev 698/452D6AA8, desc: CHECKPOINT_SHUTDOWN redo
698/46000028; tli 6; prev tli 6; fpw true; wal_level logical; xid
0:212216474; oid 6953284; multi 86737; offset 217538; oldest xid 16464388
in DB 18769; oldest multi 1 in DB 17257; oldest/newest commit timestamp
xid: 167966575/212216473; oldest running xid 0; shutdown
pg_waldump: error: error in WAL record at 698/46000028: invalid record
length at 698/460000A0: expected at least 24, got 0
On the primary summary process works fine
-rw-------. 1 postgres postgres 2340 Feb 17 10:05
000000090000069A7F9595980000069A7F9B2D08.summary
-rw-------. 1 postgres postgres 1568 Feb 17 10:35
000000090000069A7F9B2D080000069A7F9EF6C8.summary
-rw-------. 1 postgres postgres 5300 Feb 17 11:05
000000090000069A7F9EF6C80000069A7FA47288.summary
-rw-------. 1 postgres postgres 642 Feb 17 11:35
000000090000069A7FA472880000069A7FA6B820.summary
no lag between primary and secondary
Recovery on the standby is active and on time
postgres 544620 544615 0 10:50 ? 00:00:00 postgres: dev_ge:
startup recovering 000000090000069A00000080
postgres 544675 544615 0 10:50 ? 00:00:00 postgres: dev_ge:
walreceiver streaming 69A/800D0808
history files:
cat 00000007.history
1 5F4/BB000C68 no recovery target specified
2 618/9C0000A0 no recovery target specified
3 619/E70000A0 no recovery target specified
4 669/96FFFFD0 no recovery target specified
5 693/CC0000A0 no recovery target specified
6 698/460000A0 no recovery target specified
cat 00000008.history
1 5F4/BB000C68 no recovery target specified
2 618/9C0000A0 no recovery target specified
3 619/E70000A0 no recovery target specified
4 669/96FFFFD0 no recovery target specified
5 693/CC0000A0 no recovery target specified
6 698/460000A0 no recovery target specified
7 699/900000A0 no recovery target specified
cat 00000009.history
1 5F4/BB000C68 no recovery target specified
2 618/9C0000A0 no recovery target specified
3 619/E70000A0 no recovery target specified
4 669/96FFFFD0 no recovery target specified
5 693/CC0000A0 no recovery target specified
6 698/460000A0 no recovery target specified
7 699/900000A0 no recovery target specified
8 699/910000A0 no recovery target specified
In the code:
walsummarizer.c
/*
* Main loop: read xlog records one by one.
*/
while (1)
{
int block_id;
char *errormsg;
XLogRecord *record;
uint8 rmid;
ProcessWalSummarizerInterrupts();
/* We shouldn't go backward. */
Assert(summary_start_lsn <= xlogreader->EndRecPtr);
/* Now read the next record. */
record = XLogReadRecord(xlogreader, &errormsg);
if (record == NULL)
{
xlogreader.c
XLogReadRecord --> XLogReadAhead(XLogReaderState *state, bool nonblocking)
--> result = XLogDecodeNextRecord(state, nonblocking); => error
--> XLogNextRecord
/* There may be no next page if it's too small. */
if (total_len < SizeOfXLogRecord)
{
report_invalid_record(state,
"invalid record length at %X/%08X: expected at least %u, got %u",
LSN_FORMAT_ARGS(RecPtr),
(uint32) SizeOfXLogRecord, total_len);
goto err;
why walsummarizer works fine on the primary and it is stuck on the
secondary?
when the walsummarizer process was stop for a while (24h) and restarted,
it resumes, but the history is lost.
-rw-------. 1 postgres postgres 580 Feb 18 11:09
000000090000069AB9B511680000069AB9B7F260.summary
-rw-------. 1 postgres postgres 1880 Feb 18 11:09
000000090000069AB9B7F2600000069AB9BB6780.summary
-rw-------. 1 postgres postgres 530 Feb 18 11:09
000000090000069ABF2082880000069ABF23C690.summary
-rw-------. 1 postgres postgres 8582 Feb 18 11:35
000000090000069ABF23C6900000069AC01B1EB0.summary
Any clue of what happend will be very appreciated
Fabrice
