Horton

Re: pgbench bug candidate: negative "initial connection time"

Fujii Masao <masao.fujii@oss.nttdata.com>

View thread View on postgr.es
From: Fujii Masao <masao.fujii@oss.nttdata.com>
To: Yugo NAGATA <nagata@sraoss.co.jp>
Cc: Fabien COELHO <coelho@cri.ensmp.fr>, Kyotaro Horiguchi <horikyota.ntt@gmail.com>, kuroda.hayato@fujitsu.com, PostgreSQL Developers <pgsql-hackers@lists.postgresql.org>
Date: 2021-09-29T13:11:53Z
Lists: pgsql-hackers 

On 2021/09/24 11:26, Fujii Masao wrote:
> 
> 
> On 2021/09/24 7:26, Yugo NAGATA wrote:
>> That makes sense. Failures of setup connection or initial connection doesn't
>> seem 'static problems'. I rewrote this description to explain exit status 1
>> indicates also interal errors and early errors.
>>
>>    Exit status 1 indicates static problems such as invalid command-line options
>>    or internal errors which are supposed to never occur.  Early errors that occur
>>    when starting benchmark such as initial connection failures also exit with
>>    status 1.
> 
> LGTM. Barring any objection, I will commit the patch.

I extracted two changes from the patch and pushed (also back-patched) them.

The remainings are the changes of handling of initial connection or
logfile open failures. I agree to push them at least for the master.
But I'm not sure if they should be back-patched. Without these changes,
even when those failures happen, pgbench proceeds the benchmark and
reports the result. But with the changes, pgbench exits immediately in
that case. I'm not sure if there are people who expect this behavior,
but if there are, maybe we should not change it at least at stable branches.
Thought?

BTW, when logfile fails to be opened, pgbench gets stuck due to commit
aeb57af8e6. So even if we decided not to back-patch those changes,
we should improve the handling of logfile open failure, to fix the issue.

Regards,

-- 
Fujii Masao
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION

Commits

  1. pgbench: Fix typo in comment.

  2. pgbench: Improve error-handling in pgbench.

  3. pgbench: Fix handling of socket errors during benchmark.

  4. pgbench: Correct log level of message output when socket wait method fails.

  5. pgbench: Synchronize client threads.

  6. pgbench: Use common logging API