Обсуждение: Gap in sequence after fail-over

Поиск
Список
Период
Сортировка

Gap in sequence after fail-over

От
Simon Luijk
Дата:
I have two test systems with Postgresql in Hot-standby configuration. The intention
is to fail-over to the slave every so often. When ever we switch servers (fail-over)
and start inserting data, the primary ids have a gap of ~30. From the following
thread form 2001 it seems this maybe expected?

http://archives.postgresql.org/pgsql-bugs/2001-12/msg00082.php

With a Hot-standby set-up like this, that regularly switches, can I not rely on
the sequence always incrementing by 1?

Thank you
Simon

Re: Gap in sequence after fail-over

От
Simon Riggs
Дата:
On Tue, Apr 12, 2011 at 9:28 AM, Simon Luijk <simonluijk@gmail.com> wrote:
> I have two test systems with Postgresql in Hot-standby configuration. The
> intention
> is to fail-over to the slave every so often. When ever we switch servers
> (fail-over)
> and start inserting data, the primary ids have a gap of ~30. From the
> following
> thread form 2001 it seems this maybe expected?
>
> http://archives.postgresql.org/pgsql-bugs/2001-12/msg00082.php
>
> With a Hot-standby set-up like this, that regularly switches, can I not rely
> on
> the sequence always incrementing by 1?

No, you can't.

Sequences cache their values for performance. Any crash or failover
will produce the same effect.

We might change that for switchovers but it would take some effort and
would still be the same if it crashes.

--
 Simon Riggs                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services