| From: | Peter Eisentraut <peter(dot)eisentraut(at)2ndquadrant(dot)com> |
|---|---|
| To: | Petr Jelinek <petr(dot)jelinek(at)2ndquadrant(dot)com>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
| Cc: | Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: Somebody has not thought through subscription locking considerations |
| Date: | 2017-04-10 19:15:10 |
| Message-ID: | 2299a476-07ec-ac85-fd04-3ae79cb7dd94@2ndquadrant.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On 3/31/17 20:25, Petr Jelinek wrote:
> On 01/04/17 01:57, Petr Jelinek wrote:
>> That being said, looking at use-cases for SetSubscriptionRelState that's
>> basically CREATE SUBSCRIPTION, ALTER SUBSCRIPTION REFRESH and tablesync
>> worker. So the DDL thing applies to first ones as well and tablesync
>> should not be running in case the record does not exist so it's fine if
>> it fails. In terms of RemoveSubscriptionRel that's only called from
>> heap_drop_with_catalog and tablesync holds relation lock so there is no
>> way heap_drop_with_catalog will happen on the same relation. This leads
>> me to thinking that RowExclusiveLock is fine for both
>> SetSubscriptionRelState and RemoveSubscriptionRel as long as we document
>> that callers should be aware that SetSubscriptionRelState has
>> concurrency issues and fail on unique index check.
>>
>
> And a simple patch to do so. Peter do you see any problem with doing this?
committed
--
Peter Eisentraut http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Andres Freund | 2017-04-10 19:16:54 | Re: "left shift of negative value" warnings |
| Previous Message | Tom Lane | 2017-04-10 19:10:50 | Re: recent deadlock regression test failures |