On Tue, May 05, 2020 at 10:57:04AM +1200, Thomas Munro wrote:
! On Tue, May 5, 2020 at 10:13 AM Peter <pmc@xxxxxxxxxxxxxxxxxxxxxxx> wrote:
! > BTW, I would greatly appreciate if we would reconsider the need for
! > the server to read the postmaster.pid file every few seconds (probably
! > needed for something, I don't know).
! > That makes it necessary to set atime=off to get a spindown, and I
! > usually prefer to have atime=on so I can see what my stuff is
! > currently doing.
!
! That'd be this:
!
! /*
! * Once a minute, verify that postmaster.pid hasn't been removed or
! * overwritten. If it has, we force a shutdown. This avoids having
! * postmasters and child processes hanging around after their database
! * is gone, and maybe causing problems if a new database cluster is
! * created in the same place. It also provides some protection
! * against a DBA foolishly removing postmaster.pid and manually
! * starting a new postmaster. Data corruption is likely to ensue from
! * that anyway, but we can minimize the damage by aborting ASAP.
! */
! if (now - last_lockfile_recheck_time >= 1 * SECS_PER_MINUTE)
! {
! if (!RecheckDataDirLockFile())
Ah. Yes. Thanks for identifying this issue.
It is really hard to comment on this in a way that might not be
considered offensive, so lets put it that way: You all have seen
the Apollo-13 movie, so You know that a need to power-down as much
as possible may appear en-route, and in a situation where you have
lots of other issues, so what you need the least is things like
this getting in your way.
I usually take this a bit further and consider only two relevant
quality levels for software: interplanetary and interstellar - where
i.p. is the kind of stuff on which we will trust our lives during
the manned Jupiter mission. PG might well qualify for this, except
with things like the quoted; so I think such should be tagged with
appropriate compile-time switches.
Now concerning the memory leak:
That one was introduced with the work done on the GSSAPI encryption;
it goes away when setting 'hostnogssenc' in pg_hba. (And, whoever has
editor rights, there is also a little fixup needed for that switch in
pg_hba.conf; it should probably be "eleven formats".)
cheerio,
PMc
