Thanks again for your help. Your scenario of survived PID file on a
shared device is a really possible thing and I'm glad that you pointed
it out.
On other hand Zoltan on a parallel thread suggested to use "pg_ctl
status" for checking status of PostgreSQL, and that looks like a
really good idea. So please check attached patch and let me know what
you think of it.

Thanks again for your cooperation.

On Feb 19, 2008 8:20 PM, HIDEO YAMAUCHI <[EMAIL PROTECTED]> wrote:
> Hi Serge,
>
> I confirmed your patch.
>
> There was not a problem with the next case.
>
> 1)Started in one node of the resource of one PostgreSQL
> 2)Started in one node of the resource of two PostgreSQL(pgdata and port are 
> two different PostgreSQL.)
>
> But, there was a problem with the next case.
>
> 3)When there is not PID of the PID file in PostgreSQL
>
> This case, for example...
>  1.Prepare PGDATA into the shared device.
>  2.Reboot an active node.(reboot -nf)
>  3.A PID file is in condition to have stayed.
>  4.Fail-over.
>  5.In the standby node, there is a process same PID.
>  6.Fail in start of PostgreSQL.
>
> Will not the handling of pgrep be necessary to evade this problem?
>
>
> Regards,
> Hideo Yamauchi.
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>



-- 
Serge Dubrouski.

Attachment: pgsql.patch
Description: Binary data

_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to