Re: pgbench - implement strict TPC-B benchmark

Поиск
Список
Период
Сортировка
От Fabien COELHO
Тема Re: pgbench - implement strict TPC-B benchmark
Дата
Msg-id alpine.DEB.2.21.1908280713100.28828@lancre
обсуждение исходный текст
Ответ на Re: pgbench - implement strict TPC-B benchmark  (Dmitry Dolgov <9erthalion6@gmail.com>)
Ответы Re: pgbench - implement strict TPC-B benchmark  (Dmitry Dolgov <9erthalion6@gmail.com>)
Список pgsql-hackers
Hello Dmitry,

>> Well, it could be added?
>
> While doing benchmarking using different tools, including pgbench, I found it
> useful as a temporary hack to add copy freeze and maintenance_work_mem options
> (the last one not as an env variable, just as a set before, although not sure
> if it's a best idea). Is it similar to what you were talking about?

About this patch:

Concerning the --maintenance... option, ISTM that there could rather be a 
generic way to provide "set" settings, not a specific option for a 
specific parameter with a specific unit. Moreover, ISTM that it only needs 
to be set once on a connection, not per command. I'd suggest something 
like:

   --connection-initialization '...'

That would be issue when a connection is started, for any query, then the 
effect would be achieved with:

   pgbench --conn…-init… "SET maintenance_work_main TO '12MB'" ...

The --help does not say that the option expects a parameter.

Also, in you patch it is a initialization option, but the code does not 
check for that.

Concerning the freeze option:

It is also a initialization-specific option that should be checked for 
that.

The option does not make sense if

The alternative queries could be managed simply without intermediate 
variables.

Pgbench documentation is not updated.

There are no tests.

This patch should be submitted in its own thread to help manage it in the 
CF app.

-- 
Fabien.

В списке pgsql-hackers по дате отправления:

Предыдущее
От: Michael Paquier
Дата:
Сообщение: Re: REINDEX filtering in the backend
Следующее
От: Kyotaro Horiguchi
Дата:
Сообщение: Re: [HACKERS] WAL logging problem in 9.4.3?