Test restartpoints in archive recovery.
authorNoah Misch <noah@leadboat.com>
Sun, 20 Apr 2025 15:28:48 +0000 (08:28 -0700)
committerNoah Misch <noah@leadboat.com>
Sun, 20 Apr 2025 15:28:53 +0000 (08:28 -0700)
commit4399d1055627fe5ffaba47cf6ad6d07af3905e16
treecdcc92744cdd7a57724071704e0bb1a1c6b58fa0
parentd34b671a63e61d8e6002fb45f960d078e1073ea3
Test restartpoints in archive recovery.

v14 commit 1f95181b44c843729caaa688f74babe9403b5850 and its v13
equivalent caused timing-dependent failures in archive recovery, at
restartpoints.  The symptom was "invalid magic number 0000 in log
segment X, offset 0", "unexpected pageaddr X in log segment Y, offset 0"
[X < Y], or an assertion failure.  Commit
3635a0a35aafd3bfa80b7a809bc6e91ccd36606a and predecessors back-patched
v15 changes to fix that.  This test reproduces the problem
probabilistically, typically in less than 1000 iterations of the test.
Hence, buildfarm and CI runs would have surfaced enough failures to get
attention within a day.

Reported-by: Arun Thirupathi <arunth@google.com>
Discussion: https://postgr.es/m/20250306193013.36.nmisch@google.com
Backpatch-through: 13
src/test/recovery/t/045_archive_restartpoint.pl [new file with mode: 0644]