On Fri, Apr 12, 2019 at 9:07 PM Michael Paquier <michael@paquier.xyz> wrote:
>
> On Fri, Apr 12, 2019 at 07:27:44PM +0900, Masahiko Sawada wrote:
> > As far as I know there are three places where call
> > pgstat_report_wait_end before ereport(ERROR): two in twophase.c
> > andanother in copydir.c(at L199). Since we eventually call
> > pgstat_report_wait_end() in AbortTransaction(). I think that we don't
> > need to call pgstat_report_wait_end() if we're going to raise an error
> > just after that. Is that right?
>
> RecreateTwoPhaseFile() gets called in the checkpointer or the startup
> process which do not have a transaction context
Yes.
> so the wait event would not get cleaned up
But I think that's not right, I've checked the code. If the startup
process failed in that function it raises a FATAL and recovery fails,
and if checkpointer process does then it calls
pgstat_report_wait_end() in CheckpointerMain().
> It looks that 249cf070 has been rather
> inconsistent in its way of handling things.
Yeah, I think that at least handling of pgstat_report_wait_end() in
RecreateTwoPhseFile() is inconsistent in any case.
Regards,
--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center