On Wed, Mar 19, 2014 at 4:22 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
Gurjeet Singh <gurjeet@singh.im> writes: > Please find attached the patch to send transaction commit/rollback stats to > stats collector unconditionally.
That's intentional to reduce stats traffic. What kind of performance penalty does this patch impose? If the number of such transactions is large enough to create a noticeable jump in the counters, I would think that this would be a pretty darn expensive "fix".
I can't speak to the performance impact of this patch, except that it would depend on the fraction of transactions that behave this way. Perhaps the people who develop and/or aggressively use monitoring can pitch in.
Presumably, on heavily used systems these transactions would form a small fraction. On relatively idle systems these transactions may be a larger fraction but that wouldn't affect the users since the database is not under stress anyway.