tgl(at)sss(dot)pgh(dot)pa(dot)us (Tom Lane) writes:
> Chris Browne <cbbrowne(at)acm(dot)org> writes:
>> We're getting a bit of an anomaly relating to pg_stat_activity...
>> That PID has been dead for several days, but this connection is marked
>> as being open, still, after lo many days.
> This probably just means that the "backend termination" stats message
> got dropped due to heavy load. That's expected behavior in all pre-8.2
> releases: the stats system was never intended to provide
> guaranteed-exactly-correct status. PG 8.2 has reimplemented the
> pg_stat_activity view to make it more trustworthy. (The other stuff is
> still probabilistic, but being just event counters, message loss isn't
> so obvious.)
That seems a *bit* surprising; the system wasn't expected to be under
particularly heavy load during the period in question; I would have
expected "particularly light load." No matter; there may have been
some brief heavy load to cause this.
There isn't any way, short of restarting the postmaster, to get rid of
that PID, is there?
"High-level languages are a pretty good indicator that all else is
seldom equal." - Tim Bradshaw, comp.lang.lisp
In response to
pgsql-hackers by date
|Next:||From: Martijn van Oosterhout||Date: 2006-12-21 15:37:03|
|Subject: Re: column ordering, was Re: [PATCHES] Enums patch v2|
|Previous:||From: Tom Lane||Date: 2006-12-21 15:29:14|
|Subject: Re: ERROR: tuple concurrently updated |