Closed ng-pe closed 6 years ago
New test : Postgresql 10.3 : KO Posgresql with this commit : https://github.com/postgres/postgres/commit/ab2324fd468c4bc6d8b012552550ed951d97339a#diff-b69bc548b48803f3b79d575d60062556 : KO
pg_qualstats 1.0.2: no bugs
Thanks a lot for bug report and all off-list additional information. I just pushed https://github.com/powa-team/pg_qualstats/commit/68c1dacbc57de664af4ad9b6e54a4b804f2d729e which fix this issue. Can you confirm that it also fix it for you on all your servers?
For me it's okay :) I run a stress test on 2 environments and I no longer observe any crashes.
I confirm it's fixed. Good job rjuju :)
Thanks! :)
Hello,
During an important load I observe a problem... a "segmentation fault"... here is the detail : This is an installation with Powa and pg_qualstats. Installation: OS: Centos 7 (CentOS Linux release 7.4.1708 (Core)) CPU : E5620 6core / Ram: 8 GB PostgreSQL : 9.6.8 (from PGDG) Powa : 3.1.1 pg_qualstats : 1.0.3
Postgresql.conf configuration :
Powa installation :
How to reproduce the bug: Creating the test pgbench table :
script4.sql :
pg bench :
I added some NOTICE in pgqualstats to better understand the problem and I observe :
First of all i observe out of memory in the postgresql logs :
Then the "pgqs_hash" takes a rather high value at once (much higher than "pgqs_max").
This is repeated a few hundred times without the values changing...
Then after, the memory allocation requests become inconsistent...:
Until it crashed:
the coredump : bt full
When the "pg_qualstats" extension is disabled the problem does not occur.
Regards,