On 23/08/14 13:32, Burgess, Freddie wrote:
>
> Logged by: Freddie Burgess
> Email address: fburgess@radiantblue.com
> PostgreSQL version: 9.3.4
> Operating system: Red_hat Linux 6.4
> Description:
>
> Server Monitoring is reporting the following problem on our Primary
> database server.
>
> Disk % of time device was busy Servicing a transfer request > 95% and
> outside all baselines for 15 mins
>
> No backup's were executed when this error was reported
>
> Does anyone know the possible causes for this i/o bottleneck to be reported?
>
>
On the basis of that single metric it is hard to be specific. There are
a number of possibilities e.g:
- more users active at that time
- your database has grown so it is now > ram
- autovacuum cleaning up some tables
We are going to need to see a bit more detail to help any more. E.g:
- reads or writes or both?
- wait time, queue length etc
..essentially detail from sar or iostat would be helpful.
Inspection of your postgres logs for the time period might be
informative (especially if you've set log_min_statement_duration,
log_connections etc).
BTW - this sounds like the sort of thing best discussed on the
-performance list.
Cheers
Mark