| From: | "Graeme B(dot) Bell" <grb(at)skogoglandskap(dot)no> |
|---|---|
| To: | Scott Ribe <scott_ribe(at)elevated-dev(dot)com> |
| Cc: | "pgsql-admin(at)postgresql(dot)org" <pgsql-admin(at)postgresql(dot)org>, Simon Riggs <simon(at)2ndQuadrant(dot)com> |
| Subject: | Re: postgres hot-standby questions. |
| Date: | 2015-03-26 16:08:57 |
| Message-ID: | 8A0640B6-CE88-4E03-B145-0B6149DE7173@skogoglandskap.no |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-admin |
> With fully synchronous replication the performance issue is a real killer for us.
> I suppose what I'm wondering is about async normal operation, but sync operation during shutdown.
Replying to myself - but a further thought -
Assuming that you have your server configured with 1 or more hot standbys.
Are there situations where it is smart & sane to allow a controlled (slow, not emergency) shutdown to complete asynchronously without knowing if any standby got the last bits of wal?
Should a fully async shutdown with a connected standby ever happen?
You've no guarantee the standbys are up to date. Since the server is shutdown, you've no way to ask it and check without looking at the filesystem directly.
I guess the problem with "eventually consistent" is that during shutdown you know for sure that 'eventually' isn't ever going to happen unless you make it happen.
Graeme.
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Scott Ribe | 2015-03-26 16:09:08 | Re: postgres hot-standby questions. |
| Previous Message | Scott Ribe | 2015-03-26 15:58:32 | Re: postgres hot-standby questions. |