Re: incorrect docs for pgbench / skipped transactions - Mailing list pgsql-hackers

From Fabien COELHO
Subject Re: incorrect docs for pgbench / skipped transactions
Date
Msg-id alpine.DEB.2.10.1603190805170.7678@sto
Whole thread Raw
In response to incorrect docs for pgbench / skipped transactions  (Tomas Vondra <tomas.vondra@2ndquadrant.com>)
Responses Re: incorrect docs for pgbench / skipped transactions
List pgsql-hackers
Hello Tomas,

> while learning about format of the transaction log produced by pgbench, I've 
> noticed this sentence in the section describing format of the per-transaction 
> log:
>
>   The last field skipped_transactions reports the number of
>   transactions skipped because they were too far behind schedule.
>   It is only present when both options --rate and --latency-limit
>   are used.
>
> Which is wrong, because this field is only added in the aggregated log, not 
> in the per-transaction one. So we should delete this.

Indeed.

Several builtin scripts can be specified with -b as well, and -b and -f 
can be mixed, the file_no should be renamed script_no and refer to both -b 
and -f.

> It also does not explicitly explain that with --latency-limit the latency 
> column will contain "skipped" for transactions exceeding the limit latency 
> (it's only mentioned in the example output).

Ok.

> So I think we should apply the attached patch (and also backpatch it to 9.5, 
> where the --latency-limit got introduced).

Here is an updated version of your proposal: - use <literal> instead of " - use script_no and mention -b as well -
spellout skipped explanation after the sample output
 

Also, while reading the doc, I really think that the timestamp should be 
made explicit milliseconds, i.e. "123.004567" instead of "123 4567", but 
that is another question not relevant to this patch.

-- 
Fabien.

pgsql-hackers by date:

Previous
From: Andres Freund
Date:
Subject: Re: Performance degradation in commit ac1d794
Next
From: Jeff Janes
Date:
Subject: Re: Weighted Stats