shmat() fails with Not enough space on wrasse

Hello Noah,

Could you please look at sporadic test failures occurring on wrasse?

Yesterday wrasse failed pg_rewindCheck ([3]) and it's the third failure of
the same kind since July. The failure log regress_log_001_basic contains:
# Running: pg_rewind --debug --source-server port=17304 host=/home/nm/farm/tmp/sbGTlHPVwD dbname='postgres'
user=rewind_user
--target-pgdata=/export/home/nm/farm/studio64v12_6/REL_14_STABLE/pgsql.build/src/bin/pg_rewind/tmp_check/t_001_basic_primary_remote_data/pgdata
--no-sync --write-recovery-conf
pg_rewind: executing ".../inst/bin/postgres" for target server to complete crash recovery
2024-10-03 01:32:24.149 CEST [29411:1] FATAL:  shmat(id=1392509011, addr=0, flags=0x4000) failed: Not enough space
pg_rewind: error: postgres single-user mode in target cluster failed
pg_rewind: fatal: Command was: ".../inst/bin/postgres" --single -F -D
"...src/bin/pg_rewind/tmp_check/t_001_basic_primary_remote_data/pgdata" template1 < "/dev/null"
not ok 11 - pg_rewind remote

I don't know SPARC Solaris, but I've found a recommendation to increase
zone.max-locked-memory [4] in a similar situation.

As wrasse tests REL_16_STABLE and I hope it will continue doing that for
another 4 years, maybe you would find it sensible to fix this issue if it's
not more complicated...

[1] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=wrasse&dt=2024-07-22%2023%3A48%3A35
[2] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=wrasse&dt=2024-09-15%2000%3A40%3A56
[3] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=wrasse&dt=2024-10-02%2023%3A12%3A02
[4] https://www.ibm.com/support/pages/db2-activate-database-fails-sql1084c-solaris

Best regards,
Alexander