Greg Oliver <goliver at cistera.com> writes:
> I am very new to Slony-I and fairly new to Postgresql, but am attempting
> (successfully somewhat) to propogate data from 1 master to 2 remotes.
>
> It is working as expected, although CPU runs high constantly due to the
> rollbacks and commits happening constantly - Postgresql syslog follows -
> Slony log shows no data other than normal connection info.  
>
> Is this expected behaviour?  It happens every 3 seconds when NO DATA is
> even being populated in the tables.  This is with no apps even accessing
> the databases while this is going on.  Is there something I can do to
> alleviate this?  It makes our apps unusable since they are time
> sensitive and the cpu is pegged.  BTW - these are P4s with 2Gb ram
> running on RH9 w/Postgres 7.4 if it matters.

You can expect there to be some regular activity, but I wouldn't
expect this much.  And based on the queries, I don't see anything that
should be "pegging" the CPU.  The queries (which aren't surprising
ones) are all running for just a few milliseconds apiece; that doesn't
add up to the steady problem you're reporting.  The work that those
queries show off represents maybe 50ms-100ms of work every 3 seconds;
that's not enough, by itself, to make the system busy.

1.  What are the slon command line parameters?  I'm curious as to
where the 3s timeframe comes from...

2.  Try vacuuming out pg_listener on all the nodes (verbosely, so you
can see what work it does).  If there's a lot of dead space, a VACUUM
FULL won't cause any locking trouble, and should help clean things
out.  A bloated up pg_listener is a classic reason for dodgy
performance with Slony-I...
-- 
(format nil "~S@~S" "cbbrowne" "ca.afilias.info")
<http://dev6.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)