Re: [HACKERS] Broken hint bits (freeze)

On Fri, Jun 16, 2017 at 11:03 PM, Sergey Burladyan <eshkinkot@gmail.com> wrote:
> Bruce Momjian <bruce@momjian.us> writes:
>
>> On Fri, Jun 16, 2017 at 08:10:13PM +0530, Amit Kapila wrote:
>> > On Fri, Jun 16, 2017 at 7:03 AM, Sergey Burladyan <eshkinkot@gmail.com> wrote:
>> > > Bruce Momjian <bruce@momjian.us> writes:
>> > >
>> > >> !      against the old primary and standby clusters.  Verify that the
>> > >> !      <quote>Latest checkpoint location</> values match in all clusters.
>> > >
>> > > For "Log-Shipping only" standby server this cannot be satisfied, because
>> > > last WAL from master (with shutdown checkpoint) never archived.
>> > >
>> >
>> > Yeah, we have ensured that all the transactions before shutdown
>> > checkpoint got archived.  It is done in commit
>> > 2e6107cb621d003dcab0df53ac8673ea67c4e467.  However, it is not clear to
>> > me neither it is mentioned in comments why we have done it that way.
>>
>> Yes, I am confused why Sergey doesn't see that behavior.
>

The behavior reported by Sergey is what is expected i.e the last file
in which shutdown checkpoint record is written won't be archived and
there is a reason behind that.  We always perform shutdown checkpoint
(which will write shutdown checkpoint record) after requesting a xlog
switch.  Any record written after xlog switch won't be archived unless
it is so big that it consumes complete xlog segment.

> I think this last new switched WAL with shutdown checkpoint record is
> incomplete and it does not marked as *.ready in pg_xlog/archive_status/
> and not archived.
>

Yes, that's true and is expected behavior.

-- 
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com