From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Michael Paquier <michael(at)paquier(dot)xyz> |
Cc: | 임명규 <myungkyu(dot)lim(at)samsung(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>, 홍도형 <don(dot)hong(at)samsung(dot)com>, 손우성 <woosung(dot)sohn(at)samsung(dot)com> |
Subject: | Re: [Proposal] Add accumulated statistics for wait event |
Date: | 2018-07-23 13:57:25 |
Message-ID: | 15168.1532354245@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Michael Paquier <michael(at)paquier(dot)xyz> writes:
> This does not need a configure switch.
It probably is there because the OP realizes that most people wouldn't
accept having this code compiled in.
> What's the performance penalty? I am pretty sure that this is
> measurable as wait events are stored for a backend for each I/O
> operation as well, and you are calling a C routine within an inlined
> function which is designed to be light-weight, doing only a four-byte
> atomic operation.
On machines with slow gettimeofday(), I suspect the cost of this
patch would be staggering. Even with relatively fast gettimeofday,
it doesn't look acceptable for calls in hot code paths (for instance,
lwlock.c).
A bigger problem is that it breaks stuff. There are countless
calls to pgstat_report_wait_start/pgstat_report_wait_end that
assume they have no side-effects (for example, on errno) and
can never fail. I wouldn't trust GetCurrentTimestamp() for either.
If the report_wait calls can't be dropped into code with *complete*
certainty that they're safe, that's a big cost.
Why exactly is this insisting on logging timestamps and not,
say, just incrementing a counter? I think doing it like this
is almost certain to end in rejection.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Bruce Momjian | 2018-07-23 13:57:55 | Re: How can we submit code patches that implement our (pending) patents? |
Previous Message | Bruce Momjian | 2018-07-23 13:56:47 | Re: How can we submit code patches that implement our (pending) patents? |