"Magnus Hagander" <mha@sollentuna.net> writes:
>>> Actually, [commit_delay] might well hurt by introducing extra delays.
>>
>> Well, if you read the documentation, you will see that it
>> will only wait if there are at least commit_siblings other
>> transactions active. So when Bacula serializes access, there
>> will be no delays, as there is only a single transaction alive.
> Hm. Right. Well, it still won't help :-)
It could actually hurt, because nonzero time is required to go look
whether there are any other active transactions. I'm not sure whether
this overhead is enough to be measurable when there's only one backend
running, but it might be.
regards, tom lane