Improve errors related to incorrect TLI on checkpoint record replay
authorMichael Paquier <michael@paquier.xyz>
Tue, 25 Jan 2022 04:37:19 +0000 (13:37 +0900)
committerMichael Paquier <michael@paquier.xyz>
Tue, 25 Jan 2022 04:37:19 +0000 (13:37 +0900)
commit741bd3293389d451adb91190f84914a59142214f
tree4745d2e50d92325e2f8f835ac5dd052a7b088561
parent410aa248e5a883fde4832999cc9b23c7ace0f2ff
Improve errors related to incorrect TLI on checkpoint record replay

WAL replay would cause a hard crash if the timeline expected by a
XLOG_END_OF_RECOVERY, a XLOG_CHECKPOINT_ONLINE, or a
XLOG_CHECKPOINT_SHUTDOWN record is not the same as the timeline being
replayed, using the same error message for all three of them.  This
commit changes those error messages to use different wordings, adapted
to each record type, which is useful when it comes to the debugging of
an issue in this area.

Author: Amul Sul
Reviewed-by: Nathan Bossart, Robert Haas
Discussion: https://postgr.es/m/CAAJ_b97i1ZerYC_xW6o_AiDSW5n+sGi8k91Yc8KS8bKWKxjqwQ@mail.gmail.com
src/backend/access/transam/xlog.c