On Thu, Jan 20, 2022 at 05:30:44PM -0500, Tom Lane wrote:
> I wrote:
> > What I'm thinking of doing is inventing a "soft_stop" variant of
> > Cluster::stop that won't complain if pg_ctl stop fails, and then
> > having 017_shm's poll_start() call that before retrying the start
> > call.
>
> I did that, modeling the details on Cluster::start's fail_ok option.
On AIX, this pair of commits added 5-10h of runtime to 017_shm.pl.
Before 6c4a890 or 4fdbf9a:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hoverfly&dt=2022-01-19%2020%3A41%3A47
[21:14:17] t/017_shm.pl ......................... ok 7360 ms ( 0.01 usr 0.00 sys + 1.62 cusr 0.71 csys = 2.34
CPU)
After both:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hoverfly&dt=2022-01-21%2021%3A03%3A28
[07:52:31] t/017_shm.pl ......................... ok 36908794 ms ( 0.00 usr 0.00 sys + 5.16 cusr 16.04 csys = 21.20
CPU)