This post is more than 5 years old
71 Posts
0
917
what is threshold for total write latency
our SCOM monitoring software has recoreded error messages for an ESX host that is connected to a Clariion CX310. The error message states host has exceeded threshold for totalWriteLatency with value of 78ms.
what does this message mean and does it indicate that there is a problem i should look into further.
what is an acceptable threshold for totalWriteLatency?
Thanks, Mark
driskollt1
131 Posts
0
December 16th, 2009 10:00
What type of drives are you using?
FC10k, FC15k, SATA?
What RAID type? Raid5, Raid 10?
If you're using SATA RAID5 for an application that is response time sensitive, then you probably should move it to FC, maybe even RAID-10 if there are a lot of random writes to that LUN.
Did you align your partitions in your VMs?
What is your write cache size set to on your CLARiiON?
All of these things affect response time.
If you have analyzer licensed I'd gather a day's worth of data and...
I'd look at that LUN and see what its response times are
Check forced flushes/sec on the LUN
Also check % dirty pages on the SP
Look at % utilization on the RAID Group.
If you don't have analyzer licensed, then check the ESX server's performance statistics on that LUN. Look at writes per second, reads per second, total IO.
driskollt1
131 Posts
1
December 16th, 2009 10:00
And if you don't notice any impact, then maybe adjust your threshold on SCOM to 100 ms maybe.
I wouldn't worry too much about response time if your application isn't impacted. It's good to know why the response time is high though so that if it does become a problem, you can fix it quickly.