Re: pgbench progress with timestamp

From: Fabien COELHO <coelho(at)cri(dot)ensmp(dot)fr>
To: Jeff Janes <jeff(dot)janes(at)gmail(dot)com>
Cc: PostgreSQL Developers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: pgbench progress with timestamp
Date: 2015-09-08 06:25:19
Message-ID: alpine.DEB.2.10.1509080815390.17831@sto
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers


>> Use milliseconds for consistency with the '%n' log_prefix patch currently
>> submitted by Tomas Vondra in the CF.
>>
>> sh> ./pgbench -P 1 -N -T 100 -c 2
>> starting vacuum...end.
>> progress: 1.0 s, 546.0 tps, lat 3.619 ms stddev 4.426
>> progress: 2.0 s, 575.0 tps, lat 3.480 ms stddev 1.705
>>
>> sh> ./pgbench -P 1 --progress-timestamp -N -T 100 -c 2
>> starting vacuum...end.
>> progress: 1440328800.064 s, 549.0 tps, lat 3.602 ms stddev 1.698
>> progress: 1440328801.064 s, 570.0 tps, lat 3.501 ms stddev 1.704
>
> I like the idea of the timestamp. But could just always print both the
> timestamp and the elapsed time, rather than adding another switch to decide
> between them?

I agree that an option for this purpose is on the heavy side.

I'm not sure how fine it would be, though: progress lines can already be a
little bit long (under -R & -L) and I do not like wide terminal. Also, I
see --progress as a "user friendly" easy to read feature to know how
things are going during a run. A timestamp does not fall into this
category: it is pretty ugly and unreadable, so it is really out of
necessity that I'm resorting to that.

So I would rather keep the option because of the inelegance of having
timestamps printed.

--
Fabien.

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Michael Paquier 2015-09-08 07:28:02 Re: Use pg_rewind when target timeline was switched
Previous Message Thomas Munro 2015-09-08 06:25:00 Re: Horizontal scalability/sharding