Re: Re: [sqlsmith] FailedAssertion("!(XLogCtl->Insert.exclusiveBackup)", File: "xlog.c", Line: 10200)

On Tue, Aug 30, 2016 at 1:44 PM, Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
> On Tue, Aug 30, 2016 at 1:32 PM, Michael Paquier
> <michael(dot)paquier(at)gmail(dot)com> wrote:
>> On Mon, Aug 29, 2016 at 11:16 PM, Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
>>> You seem to add another entry for this patch into CommitFest.
>>> Either of them needs to be removed.
>>> https://commitfest.postgresql.org/10/698/
>>
>> Indeed. I just removed this one.
>>
>>> This patch prevents pg_stop_backup from starting while pg_start_backup
>>> is running. OTOH, we also should prevent pg_start_backup from starting
>>> until "previous" pg_stop_backup has completed? What happens if
>>> pg_start_backup starts while pg_stop_backup is running?
>>> As far as I read the current code, ISTM that there is no need to do that,
>>> but it's better to do the double check.
>>
>> I don't agree about doing that.
>
> Hmm... my previous comment was confusing. I wanted to comment "it's better
> *also for you* to do the double check whether we need to prevent pg_start_backup
> while pg_stop_backup is running or not". Your answer seems to be almost the same
> as mine, i.e., not necessary. Right?

Yes, that's not necessary to do more. In the worst case, say
pg_start_backup starts when pg_stop_backup is running. And
pg_stop_backup has decremented the backup counters, but not yet
removed the backup_label, then pg_start_backup would just choke on the
presence of the backup_label file.
--
Michael