I have a production server running postgres 8.3.11. I did a dump all
and loaded up postgres 9.0.1 on another server. On the new server,
the postgres autovacuum launcher process eats up an insane amount of
ram (I have seen 17G virt with 6.5G res). On the older version, it's
at a reasonable 9MB res after running for a month straight.
If I change the autovacuum_naptime parameter to 1, I can actually see
the memory count up in MB with top. It looks like after a while it
grows faster, with the resident memory going up about 1MB every 6-8
seconds. After about 35 minutes, the virt was 455MB and the res was
296MB.
The only change I made in the postgresql.conf (on both versions) is
setting the maximum number of connections to 1000. So all other
options are default to their particular postgres versions. Both have
the same number of databases and data. The number of databases is
134. There are no queries running on the test box with the new
version. Both are running RHEL 5 64-bit for an operating system.
Any idea how to fix this? I would really like to upgrade but this is
going to be a deal breaker.