PG 9.5.5 cores on AIX 7.1

"Abraham, Danny" <danny_abraham@bmc.com> writes:

Slow machine, high stress.
I think/hope that PG is the victim of an overstressed machine.
Has anyone faced this issue in the past?

Not to point out the obvious, but you're evidently using parallel
queries, which was a brand new thing in 9.5; and it had its share
of teething problems. I can't say whether updating to current
(9.5.22) would fix this particular issue, but it would definitely
fix a bunch of instabilities in that general area.

regards, tom lane

Customer is using 10.4 , not 9.5.5.

Does the same argument apply for upgrading to 10.12 ?

Thanks

Danny
-----Original Message-----
From: Tom Lane <tgl@sss.pgh.pa.us>
Sent: Sunday, July 19, 2020 6:04 PM
To: Abraham, Danny <danny_abraham@bmc.com>
Cc: pgsql-general@postgresql.org
Subject: [EXTERNAL] Re: PG 9.5.5 cores on AIX 7.1

"Abraham, Danny" <danny_abraham@bmc.com> writes:

Slow machine, high stress.
I think/hope that PG is the victim of an overstressed machine.
Has anyone faced this issue in the past?

Not to point out the obvious, but you're evidently using parallel queries, which was a brand new thing in 9.5; and it had its share of teething problems. I can't say whether updating to current
(9.5.22) would fix this particular issue, but it would definitely fix a bunch of instabilities in that general area.

regards, tom lane

On Sun, Jul 19, 2020 at 11:04 AM Abraham, Danny <danny_abraham@bmc.com>
wrote:

Customer is using 10.4 , not 9.5.5.

Does the same argument apply for upgrading to 10.12 ?

Running the current minor release of PostgreSQL is a pre-req when reporting
problems; moreso when it's largely impractical for someone else to
duplicate the problem.

David J.

On Sun, Jul 19, 2020 at 11:01 PM Abraham, Danny <danny_abraham@bmc.com> wrote:

Segmentation fault in _alloc_initial_pthread at 0x900000000521474
0x900000000521474 (_alloc_initial_pthread+0x1d4) e8030000 ld r0,0x0(r3)
(dbx) where
_alloc_initial_pthread(??) at 0x900000000521474
__pth_init(??) at 0x90000000051f390
uload(??, ??, ??, ??, ??, ??) at 0x9fffffff000ab70 load1(??, ??, ??, ??) at 0x900000000000b74 load(??, ??, ??) at 0x900000000001ef0 loadAndInit(??, ??, ??) at 0x90000000005b38c dlopen(??, ??) at 0x90000000009bfe0
internal_load_library(??) at 0x10014c684
RestoreLibraryState(??) at 0x10014d79c
ParallelWorkerMain(??) at 0x1000bb2d0
StartBackgroundWorker() at 0x10026cd94
maybe_start_bgworkers() at 0x10003834c
sigusr1_handler(??) at 0x10003902c
__fd_select(??, ??, ??, ??, ??) at 0x9000000001567fc
ServerLoop() at 0x1004cec90
PostmasterMain(??, ??) at 0x10003a4e8
main(??, ??) at 0x1000008f8

FWIW there was a report a decade ago that looked at least superficially similar:

/messages/by-id/09B23E7BF70425478C1330D893A722C602FEC019BD@MailSVR.invera.com