Re: [HACKERS] User-perspective knowledge about wait events

2017-09-29 Thread Michael Paquier
On Sat, Sep 30, 2017 at 4:20 AM, Schneider  wrote:
> On Tue, Sep 26, 2017 at 4:28 PM, Michael Paquier
>  wrote:
>> Gathering a set of examples on wiki page with some rough
>> analysis I think would be a good start.
>
> I don't seem to have privs to create wiki pages; can someone else make
> a page where we can begin to gather things like this?
>
> Does https://wiki.postgresql.org/wiki/Wait_Events make sense?

Yes, that looks fine. Don't yo uhave a community account? I think that
it is necessary to log in to create a new page.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers


Re: [HACKERS] User-perspective knowledge about wait events

2017-09-29 Thread Schneider
On Tue, Sep 26, 2017 at 4:28 PM, Michael Paquier
 wrote:
> Gathering a set of examples on wiki page with some rough
> analysis I think would be a good start.

I don't seem to have privs to create wiki pages; can someone else make
a page where we can begin to gather things like this?

Does https://wiki.postgresql.org/wiki/Wait_Events make sense?

-Jeremy

-- 
http://about.me/jeremy_schneider


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers


Re: [HACKERS] User-perspective knowledge about wait events

2017-09-26 Thread Michael Paquier
On Wed, Sep 27, 2017 at 3:26 AM, Schneider  wrote:
> However I think that it would be immensely helpful to start gathering
> knowledge somewhere on wait events from the user perspective.  Listing
> a few of the wait events which users will probably see most often
> along with practical suggestions of what users could further examine
> or even what they could change to make improvements on their database
> systems.

I am not sure that listing wait events individually based on the
frequency they could be seen is that helpful, but profiles for
different workloads matter. For example, Postgres has pgbench in core,
so we could look at profiles generated with its generic tests, and
look at the profiles generated every second and then append them in a
custom table. By looking at how many of them are present you can guess
by what roughly Postgres is bottlecked by, and you can conclude this
and that.

For example, seeing a lot of SyncRep events in sessions would mean
that the synchronous standby may not be following its primary very
smoothly. Gathering a set of examples on wiki page with some rough
analysis I think would be a good start.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers