Thread: pg_standby testing notes

pg_standby testing notes

From
"Merlin Moncure"
Date:
I am looking into using pg_standby (v3) in a warm standby system.  I'm
going to double check it, but same machine replication seemed to work
ok.  When I tried to do remote server log shipping however, I had some
issues.

Initial setup and launch is working ok, my archive command is:
'test ! -f /var/lib/pgsql/pitr/%f && cp %p /var/lib/pgsql/pitr/%f'

my restore command is:
pg_standby -m -t/raid/pitr/kill /raid/pitr %f %p

'/raid/pitr' is an nfs mount mounted from the primary to the standby.
I suspect the problem lies there, but I thought I'd ask.  here is my
log on the standby following a fresh pitr load:

2007-01-11 07:40:37 EST : LOG:  automatic recovery in progress
 2007-01-11 07:40:37 EST : LOG:  redo starts at 0/630000B0
 2007-01-11 07:41:37 EST : LOG:  restored log file
"000000010000000000000064" from archive
 2007-01-11 07:41:38 EST : LOG:  restored log file
"000000010000000000000065" from archive
 2007-01-11 07:41:38 EST : LOG:  restored log file
"000000010000000000000066" from archive
 2007-01-11 07:51:44 EST : LOG:  could not open file
"pg_xlog/000000010000000000000067" (log file 0, segment 103): No such
file or directory
 2007-01-11 07:51:44 EST : LOG:  redo done at 0/66FFFFC8
 2007-01-11 08:01:49 EST : PANIC:  could not open file
"pg_xlog/000000010000000000000066" (log file 0, segment 102): No such
file or directory

neither the primary or the standby are in production...the 3 restored
files were from me manufacturing work on the primary to generate
files.  Is there anything obvious I should be checking?

merlin

Re: pg_standby testing notes

From
"Simon Riggs"
Date:
On Thu, 2007-01-11 at 14:20 -0500, Merlin Moncure wrote:

> I am looking into using pg_standby (v3) in a warm standby system.  I'm
> going to double check it, but same machine replication seemed to work
> ok.  When I tried to do remote server log shipping however, I had some
> issues.
>
> Initial setup and launch is working ok, my archive command is:
> 'test ! -f /var/lib/pgsql/pitr/%f && cp %p /var/lib/pgsql/pitr/%f'
>
> my restore command is:
> pg_standby -m -t/raid/pitr/kill /raid/pitr %f %p
>
> '/raid/pitr' is an nfs mount mounted from the primary to the standby.
> I suspect the problem lies there, but I thought I'd ask.  here is my
> log on the standby following a fresh pitr load:
>
> 2007-01-11 07:40:37 EST : LOG:  automatic recovery in progress
>  2007-01-11 07:40:37 EST : LOG:  redo starts at 0/630000B0
>  2007-01-11 07:41:37 EST : LOG:  restored log file
> "000000010000000000000064" from archive
>  2007-01-11 07:41:38 EST : LOG:  restored log file
> "000000010000000000000065" from archive
>  2007-01-11 07:41:38 EST : LOG:  restored log file
> "000000010000000000000066" from archive
>  2007-01-11 07:51:44 EST : LOG:  could not open file
> "pg_xlog/000000010000000000000067" (log file 0, segment 103): No such
> file or directory
>  2007-01-11 07:51:44 EST : LOG:  redo done at 0/66FFFFC8
>  2007-01-11 08:01:49 EST : PANIC:  could not open file
> "pg_xlog/000000010000000000000066" (log file 0, segment 102): No such
> file or directory
>
> neither the primary or the standby are in production...the 3 restored
> files were from me manufacturing work on the primary to generate
> files.  Is there anything obvious I should be checking?

All log lines are normal, apart from the PANIC...

This looks familiar to me; I see from my notes that I wanted to exclude
-m option, but I left it in. On review, I can't see how -m would work at
all without a (minor) change to the backend - it works... apart from the
very last file request.

Can you try -l and see if that works instead? It should perform the
same, roughly.

Thanks for your feedback.

--
  Simon Riggs
  EnterpriseDB   http://www.enterprisedb.com