From: | James Sebastian <james(dot)sebastian(at)gmail(dot)com> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | pgsql-novice(at)postgresql(dot)org |
Subject: | Re: Possible Corrputed shared memory |
Date: | 2015-08-02 03:11:13 |
Message-ID: | CA+ehAmFA4VOg61MxujJ15zk1A_+GGcsSoTa8pY4EPA3bDDkSMw@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-novice |
>> [ scratches head... ] It should certainly not have taken very long to
> >> replay 10 WAL segments worth of data. I surmise that the problems
> >> you were having before the shutdown were worse than you thought, ie
> >> checkpoints were failing to complete, probably due to a persistent
> >> I/O error, so that there was a whole lot more than normal to replay
> >> after the last successful checkpoint. Is there any evidence of such
> >> distress in the postmaster log?
>
> > We had very slow application performance and many hanging threads as per
> > pgadmin -> server status
> > Also logs had the following which also indicating probably high I/O (as
> per
> > google search results)
>
> > 2015-07-30 10:10:21 IST WARNING: pgstat wait timeout
> > 2015-07-30 10:12:21 IST WARNING: pgstat wait timeout
>
> Well, those might mean problems with the stats collector subprocess, but
> that's pretty noncritical; it would certainly not have prevented
> checkpoints from completing. No other unexplained log entries?
>
>
None. I have default log configuration as given by Ubuntu as default.
Probably I would need to increase them to get some more detailed level. I
will do some reading on them.
As of now, logging_collector and other log related are commented out, so
whatever is default is functioning and logging to postgresql-9.1-main.log
If you have some suggestions, please let me know.
One area of concern, I am realising now which might have contributed to
this is below.
I am using postgres as backend db of a java based application over tomcat.
JVM is allocated 2048 MB memory heap size from tomcat.
But shared_buffers in postgres was at the default of 32m. I increased them
to 144 M and increased shmmax to 320 M. This machine has 8 GB RAM and used
as Database and Tomcat/application server. I can allocate more memory, but
my application side colleagues like to know why we keep high and how to
figure out instead of blindly allocating a high memory.
I am not even sure how to figure out what should be my buffers.
I now kept
effective_cache_size = 256MB
max_connections = 100
work_mem = 10MB
and all others are default provided by postgres 9.1 in Ubuntu 12.04
Let me know any of them looks bad and worse.
Regards
James
From | Date | Subject | |
---|---|---|---|
Next Message | Wei Shan | 2015-08-03 07:17:07 | Re: Possible Corrputed shared memory |
Previous Message | Tom Lane | 2015-08-01 15:03:41 | Re: Possible Corrputed shared memory |