From: | Michael Paquier <michael(dot)paquier(at)gmail(dot)com> |
---|---|
To: | Andres Freund <andres(at)anarazel(dot)de> |
Cc: | Robert Haas <robertmhaas(at)gmail(dot)com>, Jeff Janes <jeff(dot)janes(at)gmail(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Re: BUG #13685: Archiving while idle every archive_timeout with wal_level hot_standby |
Date: | 2015-11-07 06:54:08 |
Message-ID: | CAB7nPqQDZkhjZeR4f1Bfn=rLHD_auD57ZLUs10s=CwuCSybENQ@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-bugs pgsql-hackers |
On Sat, Nov 7, 2015 at 1:52 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> On 2015-11-06 11:42:56 -0500, Robert Haas wrote:
>> On Fri, Nov 6, 2015 at 2:47 AM, Michael Paquier
>> <michael(dot)paquier(at)gmail(dot)com> wrote:
>> > I have as well thought a bit about adding a space-related constraint
>> > on the standby snapshot generated by the bgwriter, so as to not rely
>> > entirely on the interval of 15s. I finished with the attached that
>> > uses a check based on CheckPointSegments / 8 to be sure that at least
>> > this number of segments has been generated since the last checkpoint
>> > before logging a new snapshot. I guess that's less brittle than the
>> > last patch. Thoughts?
>>
>> I can't see why that would be a good idea. My understanding is that
>> the logical decoding code needs to get those messages pretty
>> regularly, and I don't see why that need would be reduced on systems
>> where CheckPointSegments is large.
>
> Precisely.
>
>
> What I'm thinking of right now is a marker somewhere in shared memory,
> that tells whether anything worthwhile has happened since the last
> determination of the redo pointer. Where standby snapshots don't
> count. That seems like it'd be to maintain going forward than doing
> precise size calculations like CreateCheckPoint() already does, and
> would additionally need to handle its own standby snapshot, not to speak
> of the background ones.
I thought about something like that at some point by saving a minimum
activity pointer in XLogCtl, updated each time a segment was forcibly
switched or after inserting a checkpoint record. Then the bgwriter
looked at if the current insert position matched this minimum activity
pointer, skipping LogStandbySnapshot if both positions match. Does
this match your line of thoughts?
> Seems like it'd be doable in ReserveXLogInsertLocation().
> Whether it's actually worthwhile I'm not all that sure tho.
I am not sure doing extra work there is a good idea. There are already
5 instructions within the spinlock section, and this code path is
already high in many profiles for busy systems.
--
Michael
From | Date | Subject | |
---|---|---|---|
Next Message | Michael Paquier | 2015-11-07 07:09:57 | Re: BUG #13755: pgwin32_is_service not checking if SECURITY_SERVICE_SID is disabled |
Previous Message | David Gould | 2015-11-07 00:46:49 | Re: BUG #13750: Autovacuum slows down with large numbers of tables. More workers makes it slower. |
From | Date | Subject | |
---|---|---|---|
Next Message | Michael Paquier | 2015-11-07 07:09:57 | Re: BUG #13755: pgwin32_is_service not checking if SECURITY_SERVICE_SID is disabled |
Previous Message | Robert Haas | 2015-11-07 04:03:31 | Re: ALTER INDEX...SET tab completion |