Re: pgbench: Skipping the creating primary keys after initialization

On Fri, Sep 1, 2017 at 4:42 PM, Fabien COELHO <coelho(at)cri(dot)ensmp(dot)fr> wrote:
>
> Hello Masahiko-san,
>
> Patch applies and compiles.
>
> One bug found, and some minor points again. Sorry for this hopefully last
> iteration... I'm kind of an iterative person...
>
> I've generated the doc to look a it.
>
> Short option "-I" does not use a "=", it should be "-I
> custom_init_commands".
>
> Also maybe it would look nicer and clearer if the short mnemonic was outside
> the literal, that is with:
>
> <term><literal>c</> (cleanup)</>
>
> instead of:
>
> <term><literal>c (cleanup)</></>
>
> But this is debatable. Do it the way you think is best.
>
> Command "g" does not work after "f", something I had not tested before:
>
> ./pgbench -i -I ctvpfg
> cleaning up...
> creating tables...
> vacuum...
> set primary keys...
> set foreign keys...
> ERROR: cannot truncate a table referenced in a foreign key constraint
> DETAIL: Table "pgbench_history" references "pgbench_accounts".
> HINT: Truncate table "pgbench_history" at the same time, or use TRUNCATE
> ... CASCADE.
>
> I think it should work. It probably just mean to TRUNCATE all tables as one
> command, or add the suggested CASCADE. I would favor the first option.
>
> I'm wondering whether this truncation should be yet another available
> command? Hmmm... maybe not.

Currently TRUNCATE pgbench_accounts command is executed within a
transaction started immediately before it. If we move it out of the
transaction, the table data will be truncated even if the copying data
failed. Maybe we can do TRUNCATE pgbench_accounts, pgbench_history
instead. Thought?

Regards,

--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center