Jeff Janes wrote:
> This patch seems to solve a problem I've also been having with non-existent
> "pg_multixact/members/13D35" files in my testing of torn page write and fk
> locks again recent 9.4. However, sometimes I've found the problem even
> before multi wrap around occured, and my reading of this thread suggests
> that the current issue would not show up in that case, so I don't know how
> much comfort to take in the knowledge that the problem seems to have gone
> away--as they say, things that go away by themselves without explanation
> also tend to come back by themselves.
I think this problem is rather related to the code touched by commit
722acf51a0d074d19782ad7e97ebe3fdb63fbb87, which is about removing
unneeded files in pg_multixact/members. I checked your files, and it
seems that pg_multixact/members files that should still exist are gone;
for instance, pg_controldata says that the oldest multi is 2586314407,
but if I run
select * from pg_get_multixact_members('that value')
I get an error saying that pg_multixact/members/218B does not exist.
It seems that there are a lot of files that were improperly removed.
I expect that existing files should be sequential, but there are a lot
of holes in a listing of pg_multixact/members.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services