Re: Re: Patch to send transaction commit/rollback stats to the stats collector unconditionally.

On Tue, Jul 1, 2014 at 7:32 PM, Gurjeet Singh <gurjeet@singh.im> wrote:
> On Tue, Jul 1, 2014 at 10:05 AM, Kevin Grittner <kgrittn@ymail.com> wrote:
>> Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>
>>> If we're going to do it like this, then I think the force flag
>>> should be considered to do nothing except override the clock
>>> check, which probably means it shouldn't be tested in the initial
>>> if() at all.
>>
>> That makes sense, and is easily done.
>
> Attached is the patch to save you a few key strokes :)
>
>> The only question left is
>> how far back to take the patch.  I'm inclined to only apply it to
>> master and 9.4.  Does anyone think otherwise?
>
> Considering this as a bug-fix, I'd vote for it to be applied to all
> supported releases. But since this may cause unforeseen performance
> penalty, I think it should be applied only as far back as the
> introduction of PGSTAT_STAT_INTERVAL throttle.
>
> The throttle was implemeted in 641912b, which AFAICT was part of 8.3.
> So I guess all the supported releases it is.

I'll vote for master-only.  I view this as a behavior change, and it
isn't nice to spring those on people in minor releases.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company