Postgres Pro
Downloads
Home   >   mailing lists

Thread: Logging auto_explain outputs to another log file

Logging auto_explain outputs to another log file

From
Devrim GÜNDÜZ
Date:
Hi,

Is $SUBJECT possible? If not, do you think it would be worth
implementing this?

If this parameter is not used appropriately, it may fill up logs --
that's why I am asking this question.

Regards,
--
Devrim GÜNDÜZ, RHCE
devrim~gunduz.org, devrim~PostgreSQL.org, devrim.gunduz~linux.org.tr                  http://www.gunduz.org

Re: Logging auto_explain outputs to another log file

From
Tom Lane
Date:
Devrim GÜNDÜZ <devrim@gunduz.org> writes:
> Is $SUBJECT possible? If not, do you think it would be worth
> implementing this?

No, and no.  The feature isn't even in core; it can hardly qualify as
something that should drive a massive overhaul of the elog
infrastructure.  Which is what this would take.
        regards, tom lane

Re: Logging auto_explain outputs to another log file

From
Andrew Dunstan
Date:

Tom Lane wrote:
> Devrim GÜNDÜZ <devrim@gunduz.org> writes:
>   
>> Is $SUBJECT possible? If not, do you think it would be worth
>> implementing this?
>>     
>
> No, and no.  The feature isn't even in core; it can hardly qualify as
> something that should drive a massive overhaul of the elog
> infrastructure.  Which is what this would take.
>
>             
>   

Well, it might be possible to build some sort of splitting facility 
(regex based?) into the logging collector without having to change the 
rest of the logging infrastructure.

But there are already good log splitting tools for some varieties of 
syslog, and like Tom I suspect using any effort in this direction on our 
part is probably not worth it.

cheers

andrew

Re: Logging auto_explain outputs to another log file

From
Robert Treat
Date:
On Sunday 23 November 2008 15:50:09 Andrew Dunstan wrote:
> Tom Lane wrote:
> > Devrim GÜNDÜZ <devrim@gunduz.org> writes:
> >> Is $SUBJECT possible? If not, do you think it would be worth
> >> implementing this?
> >
> > No, and no.  The feature isn't even in core; it can hardly qualify as
> > something that should drive a massive overhaul of the elog
> > infrastructure.  Which is what this would take.
>
> Well, it might be possible to build some sort of splitting facility
> (regex based?) into the logging collector without having to change the
> rest of the logging infrastructure.
>
> But there are already good log splitting tools for some varieties of
> syslog, and like Tom I suspect using any effort in this direction on our
> part is probably not worth it.
>

How would you folks feel about adding a dtrace probe to look for this? I 
haven't exactly worked out where/how this would be put, but it would allow 
for easily tracking these via dtrace if we had one. 

-- 
Robert Treat
Conjecture: http://www.xzilla.net
Consulting: http://www.omniti.com