Re: trace_checkpoint parameter patch

Satoshi Nagayasu wrote:
> Hi all,
>
> Here is a brand new patch to log a checkpointing load information
> to tune the bgwriter parameter.
>
> When you enable trace_checkpoint parameter and process
> the checkpoint, the bgwriter logs the number of
> flushed buffer pages and the elapsed time.

I don't think it needs to be in a TRACE_CHECKPOINT ifdef ... (and even
if it does, I think this would fail to compile with that one disabled).


> -----------------------------------------------------------
> {61}snaga@artemis:~/pgsql-snapshot% ./bin/psql -c 'checkpoint' pgbench
> LOG:  CheckPoint: 1522/3072 buffer(s) flushed. CPU 0.03s/0.00u sec elapsed 1.57 sec
> CHECKPOINT
> {62}snaga@artemis:~/pgsql-snapshot%
> -----------------------------------------------------------
>
> I think this information is useful to configure
> the bgwriter parameter to reduce the performance impact
> of the checkpointing.
>
> Any suggestion or comments?

Well, more I/O numbers would be more interesting than CPU stats ...

--
Alvaro Herrera                 http://www.amazon.com/gp/registry/DXLWNGRJD34J
"Everybody understands Mickey Mouse. Few understand Hermann Hesse.
Hardly anybody understands Einstein. And nobody understands Emperor Norton."