On Fri, 2005-03-04 at 11:38, Lee Wu wrote:
> Greetings,
>
>
>
> To best of my knowledge, nobody issued pg_ctl stop –m fast while
>
> we experienced PG was trying to shutdown unsuccessfully:
>
>
>
> *** Everything seems normal before this point ***
>
> Mar 3 21:07:55 myhost postgres[630]: [24] LOG: recycled transaction
> log file 00001ABC0000006D
>
> Mar 3 21:08:57 myhost postgres[5158]: [1] LOG: fast shutdown request
>
> Mar 3 21:08:57 myhost postgres[5158]: [2] LOG: aborting any active
> transactions
>
> Mar 3 21:08:57 myhost postgres[14292]: [1] FATAL: This connection
> has been terminated by the administrator.
>
> *** A lot of “This connection has been terminated by the
> administrator.” taken out here ***
>
> Mar 3 21:08:57 myhost postgres[15912]: [3] FATAL: The database
> system is shutting down
>
> *** A lot the same message as last line till ***
>
> Mar 3 22:10:39 myhost postgres[1840]: [3] FATAL: The database system
> is shutting down
>
>
>
> Then I used “–m immediate”:
>
> Mar 3 22:10:39 myhost postgres[5158]: [3] LOG: immediate shutdown
> request
>
> Mar 3 22:11:07 myhost postgres[1918]: [1] LOG: database system was
> interrupted at 2005-03-03 21:08:53 MST
>
> Mar 3 22:11:07 myhost postgres[1918]: [2] LOG: checkpoint record is
> at 1ABC/929DFCB4
>
> Mar 3 22:11:07 myhost postgres[1918]: [3] LOG: redo record is at
> 1ABC/9044F3B4; undo record is at 0/0; shutdown FALSE
>
> Mar 3 22:11:07 myhost postgres[1918]: [4] LOG: next transaction id:
> 1090597972; next oid: 4231750970
>
> Mar 3 22:11:07 myhost postgres[1918]: [5] LOG: database system was
> not properly shut down; automatic recovery in progress
>
> Mar 3 22:11:07 myhost postgres[1918]: [6] LOG: redo starts at
> 1ABC/9044F3B4
>
> Mar 3 22:11:07 myhost postgres[1919]: [1] FATAL: The database system
> is starting up
>
> *** PG was up after a lot of the above message ***
>
>
>
> PG version is 7.3.2 on Red Hat Enterprise Linux ES release 3 (Taroon
> Update 4)
>
> I know it is old, but anyone can smell why PG shutdown itself?
This sounds like postgresql ran up against the Out Of Memory killer in
the linux kernel, but I thought it killed things with -9, which would
result in an immediate shutdown. Maybe a ulimit issue? Those are the
two most common causes of postgresql premature shutdown.