ctl_cyrusdb causing high load spikes

[Paul Boven]

L.S.,

We're running Cyrus IMAP/POP server for apx. 30 users, which has been in 
production since december and we've slowly been migrating users to it.
The system is a SunFire V120, Solaris 9, and the Cyrus version is 
2.1.10. We use Berkley-DB 4.1.24 with it.

Cyrus has been configured like this:
./configure  --with-perl=/bin/perl --with-sasl=/usr/local 
--with-cyrus-prefix=/usr/local/cyrus 
--with-dbdir=/usr/local/BerkeleyDB.4.1 --with-auth=unix

Now if you take a look at the three graphs located in 
http://home.sara.nl/~boven/plaatjes you will notice the repeating spikes 
when ctl_cyrusdb is run. These grow bigger and bigger (ctl_cyrusdb takes 
longer to complete while using all CPU resources), and the 
database-checkpointing is causing a much higher overall load on the 
system than the actual mail-filtering and serving. And then suddenly the 
  checkpointing is done within seconds again, and the cycle repeats itself.

I would appreciate any hints on what is going wrong here (if anything), 
and how the performance of the checkpointing could be improved without 
losing the ability to recover the database.

Regards, Paul Boven.