heavily IO bound - how to diagnose? :-(

hi,

got a performance issue on 280R running solaris8 with S1 attached - and it seems to be disk issue :-(

sar constantly shows %wio as 30-60, disks become busy up to 100% sometimes, and i only get speed of 1meg/s when doing file copy from disk to disk.

are there any ways to diagnose this problem?

best regards,

ranger.

[363 byte] By [ranger] at [2007-11-25 23:39:21]
# 1
Take a gander through the document off of this guy's web page: http://users.tpg.com.au/adsln4yb/Perf/solarismetrics.htmlIt might help identify which process is binding up your system.
Chesapeake at 2007-7-5 18:23:01 > top of Java-index,General,Talk to the Sysop...
# 2

* Do you get any messages in /var/adm/messages, about defective hard disks?

* Are you using a volume manager? If so, does it give you any errors about the status of your disks?

* Does "format" return any errors about your disks?

* How is the S1 connected to your 280? Was everything done by the book?

1 meg a second really is dreadfully slow!

Cailin_Coilleach at 2007-7-5 18:23:01 > top of Java-index,General,Talk to the Sysop...
# 3

run "iostat -nE" and see if any particular disk is showing soft/hard/tran errors (and by running this periodically, you can see if this count is increasing). That way you can identify whether you have a disk problem.

Also, like someone mentioned, check your /var/adm/messages log files and dmesg to see whether there are "scsi" and disk related error messages.

HTH

implicate_order at 2007-7-5 18:23:01 > top of Java-index,General,Talk to the Sysop...