|Main Archive Page > Month Archives > drbd-user archives|
I have an OLTP application running on a 2 node cluster on top of DRBD.
( App: CentOS Linux --> MySql --> Perl --> Apache )
( Disk: Disk Partition --> LV --> DRBD --> Mount point )
( Cluster: DRBD <-- primary/primary --> DRBD )
Before I moved to this DRBD cluster setup, my response-time-graph was
very consistent and stable (flatline).
After I moved to DRBD, the response-time-graph became inconsistent
with peaks that don't make sense. Eg. typical response time would be
0.2 seconds. Now I have responses of, for instance 3 or more seconds
every now and then). The cluster hardware have more CPU's, more
memory, faster everything, and yet the response-times are slower and
less consistent than my "no DRBD" setup. It's been running in
production with DRBD for several months. Everything seems stable,
other than the inconsistent performance.
Could this be a misconfiguration on my part (or lack of tuning), or is
this normal? The server load is under 40%. The disks aren't very
busy either. The response-time-graph is fed by a few queries on the
database across important tables.
Any guidance or help will be MUCHOS appreciated :)
Thanks and Regards,
Peet de Vos
082 629 9915
Fax: +27 12 803 5084 - P.O.Box 75250, Lynnwood Ridge, 0040, Pretoria,
drbd-user mailing list