#55 om_mysql system utilisation problem
I have a agent-server setup (msyslogd agent station
and mysql server), where syslog messages received by
agent are stored on the server.
agent is suse 9.0 distro i386 (1GB ram P4 2.8 ghz;80GB)
kern 2.4.21;mysql 4.0.15
server is suse 9.1 distro i386 (2GB P4 2.8;150GB) kern
2.6.3
Msyslog ver 1.08g
The syslog database got cca 50mil entries and is cca
6GB big.
when I perform some operations on the database
(select/insert/delete) it takes a bit time to do that and
the table is locked for inserts during that time.
Problem:
With queries taking over cca 30min it happens that cron
on agent machines gets multiplied (forked) and within
cca 20 minutes loads nr. of processes pretty good and
then kernel starts killing random processes.
Strange enough it is only on agent machines and only
when it table is locked for longer time. Delayed inserts
helps to extend the time when this behavior happens.
All system variables seem to be normal (even the load)
except number of processes. PS at that point of time
shows may cron processes (whether there are any
entries in crontab or not).
This problem is reproducible (always under same
circumstances) and also happens on redhat machine 9.x
too. latest patches from distros applied, but doesnt help
either.
And whole setup works if the syslog database is not
being processed (storage works).