On 9/19/14 1:54 AM, Huang, Suya wrote:
> How to compare buffers_backend against buffers_checkpoint?
This is a very general question, but generally you want buffers_backend
to be as small as possible.
> Below is some statistics of check point activities happened on the day while backend is loading data:
>
> Day Hour Written buffers Write time Sync time Total time
> 06 107,338 1,475.583s 3.142s 1,478.828s
> 07 601,672 1,041.482s 348.215s 1,390.153s
> 08 3,613 361.422s 0.053s 361.535s
> 09 1,006,704 1,456.898s 14.501s 1,471.545s
> 10 1,116,463 1,987.896s 6.164s 1,995.11s
>
> Can we tell something about the configuration?
If this is while loading data, there isn't much to say. I'd be worried
about the 348s sync time if this were during normal transaction
processing load.
