From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Andrew Dunstan <andrew(at)dunslane(dot)net> |
Cc: | pgsql-hackers(at)postgreSQL(dot)org |
Subject: | Re: pg_dump and dependencies and --section ... it's a mess |
Date: | 2012-06-22 00:58:32 |
Message-ID: | 25983.1340326712@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
> On 06/21/2012 07:43 PM, Tom Lane wrote:
>> I didn't understand that then, and I still don't.
> If something else holds a lock on the table (e.g. another CREATE INDEX)
> the ALTER TABLE will block until it's done, waiting for an ACCESS
> EXCLUSIVE lock. The whole method of operation of parallel restore is
> that we are not supposed to start items that might be blocked by
> currently running operations.
Oh, I see --- Andrew Hammond's complaint involves *three* index
creations on the same table (one running; a second one completed except
for the ALTER TABLE CLUSTER step attached to it, which is blocked behind
the running index creation; and a third blocked behind the ALTER TABLE
CLUSTER). I had misread it to imply that only two indexes were enough
to manifest the problem.
Yeah, we should break out the ALTER TABLE CLUSTER step as a separate
TOC item to fix this.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Shigeru HANADA | 2012-06-22 01:06:53 | Re: [PATCH] Allow breaking out of hung connection attempts |
Previous Message | Andrew Dunstan | 2012-06-22 00:32:14 | Re: pg_dump and dependencies and --section ... it's a mess |