ctl_cyrusdb I/O requirements

Henrique de Moraes Holschuh hmh at debian.org
Sun Oct 4 12:59:58 EDT 2009


On Sun, 04 Oct 2009, Anthony Tibbs wrote:
> number of years now.  All was well until last night when performance
> went down the tubes and I started seeing high I/O wait times.  I'm
> pretty sure this is a problem on the hosting end, but in trying to
> diagnose the source of the I/O congestion, the closest I could come

Rebuilding RAID arrays?  "sharing" the spindles with some other virtual
server whose workload seeks the disk around like crazy? 

> to answer was that ctl_cyrusdb is reading/writing about 15-20 4096b
> blocks (from the /var/imap/db/* files) per second.    This amounts
> to read/write throughput in the 250kb/second range.

Even a laptop HD can do > 200 blocks/s.  If you're *sure* it is not the
storage/disk going bonkers, or your provider giving you the shaft on the
sharing of IOPS to the real storage for your virtual server, it means that
box (all virtual servers AND the hypervisor) needs a switft boot to the
head because the IO scheduler went insane.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh


More information about the Info-cyrus mailing list