I have upgraded my dual processor AMD64 server running FreeBSD 5.2.1 to
PostgreSQL 7.4.3 and decided to being making use of pg_autovacuum in
order to avoid any serious problems with database integrity. This of
course meant that I had to enable row-level logging and have the new
daemon run int he background.

On the plus side, things seem to be running a bit more smoothly as it
irons out the operations of certain tables, even tables that I never
thought saw that much traffic overall. However there seems to be an
issue with load averages.

Without row level logging I would rarely see the server peak above a
load average of 1.00 with 125 concurrent connections made through web
clients. Now with this new setup I see load averages peak up to 11.75
during busy hours, meaning that overall efficiency has gone down.

Has anyone else noticed this type of issue and should I be worries about
it? Load average in my understanding means that if you have an average
of 13.00 you essentially have 13 processes waiting for the processors
time, the higher the average the slower the response and in extreme
cases your chances of crashing increase.

So again should I be worried?

I also noticed that from 7.4.2 to 7.4.3 memory utilization changed
somewhat, previous versions would leave about 600 MB free of my 2GB RAM
available while 7.4.3 will average 100 to 10 MB free, indicating that
it's using more what's available then before. Is this expected as well?

Martin Foster
Creator/Designer Ethereal Realms
martin (AT) ethereal-realms (DOT) org