[cobalt-users] Logrotate causes server problems

["Gavin Nelmes-Crocker" <cobalt@xxxxxxxxxxxxxxxx>]

Hi

I have started to see a nasty trend on one of the servers I look after, its
fully patched and has 512Mb large disk etc.

When logrotate starts at 4am it starts processing but then it seems to never
stop and eats cpu cycles, and I then have to go in and kill the processes
which then seem to build up over a few days until a reboot clears it all up
and we start again.  I'm always wary of doing a reboot when I am hundreds of
miles from the server but worse of course is that I am seeing the access
file increase in size

-rw-r--r--    1 root     root     33858825 May 12 22:27 access

I have temporarily moved logrotate out of cron so that we can look to see
what is happening but then last night/this morning something happened and
the whole lot stopped responding except to ping the only solution was for
the data centre to reboot - they said they had seen this before something
about load averages and kflushd not doing its job.

First is anyone else seeing this and secondly has anyone solved it or got
any pointers.

Thanks

Regards

Gavin