Hello,
I came across this issue while I was tweaking a TAP test with Ashwin for
this thread [1].
We noticed that while the startup process waits for a recovery delay, it does
not respect changes to the recovery_min_apply_delay GUC. So:
// Standby is waiting on recoveryWakeupLatch with a large
recovery_min_apply_delay value
node_standby->safe_psql('postgres', 'ALTER SYSTEM SET
recovery_min_apply_delay TO 0;');
node_standby->reload;
// Standby is still waiting, it does not proceed until the old timeout
is elapsed.
I attached a small patch to fix this. It makes it so that
recovery_min_apply_delay is reconsulted in the loop to redetermine the wait
interval. This makes it similar to the loop in CheckpointerMain, where
CheckPointTimeout is consulted after interrupts are handled:
if (elapsed_secs >= CheckPointTimeout)
continue; /* no sleep for us ... */
I have also added a test to 005_replay_delay.pl.
Regards,
Soumyadeep(VMware)
[1] https://www.postgresql.org/message-id/flat/CANXE4Tc3FNvZ_xAimempJWv_RH9pCvsZH7Yq93o1VuNLjUT-mQ%40mail.gmail.com