From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Heikki Linnakangas <hlinnakangas(at)vmware(dot)com> |
Cc: | PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Specifying both recovery_target_xid and recovery_target_time |
Date: | 2014-01-08 15:43:13 |
Message-ID: | 6321.1389195793@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Heikki Linnakangas <hlinnakangas(at)vmware(dot)com> writes:
> The docs say:
>> At most one of recovery_target_time, recovery_target_name or recovery_target_xid can be specified
> However, the code actually allows them all to be specified at the same time:
>> else if (strcmp(item->name, "recovery_target_name") == 0)
>> {
>> /*
>> * if recovery_target_xid specified, then this overrides
>> * recovery_target_name
>> */
>> if (recoveryTarget == RECOVERY_TARGET_XID)
>> continue;
>> recoveryTarget = RECOVERY_TARGET_NAME;
> The precedence is XID, time, name.
> I think the documented behavior would make more sense, ie. throw an
> error if you try to specify multiple targets. Anyone remember if that
> was intentional? Any objections to change the code to match the docs, in
> master?
Hm. I can see potential uses for specifying more than one if the behavior
were "OR", that is stop as soon as any of the specified conditions is
satisfied.
It looks like the actual behavior is to randomly choose one specified
mode and ignore the others, which I concur is bogus. But maybe we
should try to do something useful instead of just throwing an error.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Andres Freund | 2014-01-08 16:04:36 | Re: Changeset Extraction Interfaces |
Previous Message | Heikki Linnakangas | 2014-01-08 15:38:00 | Specifying both recovery_target_xid and recovery_target_time |