Hello,
> > After all, pmState changes to PM_NO_CHILDREN via PM_WAIT_DEAD_END
> > by SIGCHLDs from non-significant processes, then CancelBackup().
>
> Judging from what was being said on the thread, it seems that running
> CancelBackup() after an immediate shutdown is better than not doing it,
> correct?
Agreed. I like that behavior:) It removes backup_label at
immediate shutdown and (altough I didn't see by myself but as far
as I saw PostmasterStateMachine) it would skip shutdown
checkponit.
> > Focusing on the point described above, the small patch below
> > rewinds the behavior back to 9.3 and before but I don't know the
> > appropriateness in regard to the intention of the patch.
>
> I see. Obviously your patch would, in effect, revert 82233ce7ea
> completely, which is not something we want. I think if we want to go
> back to the previous behavior of not stopping the backup, some other
> method should be used.
As I mentioned above, I don't want to rewind 9.4's behavior back
to that of previous ones.
What I hope to be realized for now is '-b'(provisional optname)
of pg_resetxlog for at least versions which would fall into this
problem. What do you think about this maybe 'New Feature' but has
meaning practically only for older versions?
Of course I agree with that 'you should erase the backup_label
just after master has crashed' is the most clean and sane way to
*avoid* the situation but the penalty seems a bit too large for
the mistake.
regards,
--
Kyotaro Horiguchi
NTT Open Source Software Center