2009-06-17 07:59 AM
In PA - Volume Latency View, two questions:
1) Why does NFS Latency by Optype contain different data than Overall Latency by Optype? The volume only contains data accessed through NFS. The latency in milliseconds is off by at least factor of 10.
2). In the NFS Latency by Optype view what is the difference between nfs_latency and the other counters? I'm aware of the differences between read/write/other. Is nfs_latency meant to be a sum of those three? If so, it's off by almost a factor of 100 (sometimes almost an order of magnitude).
Solved! SEE THE SOLUTION
2009-06-17 10:57 AM
Providing a little more detail: DFM 3.7.1 (this has been continuously upgraded since 3.4) on SLES 9 sp3. This behavior is true regardless of the volume (or storage system) that is selected.
2009-06-17 10:46 PM
nfs_latency is calculated as average for of all the ops, so it would be average of read, write and other latencies. Discrepancy in the value could be due to bug in OpsMgr 3.7.1 which is fixed in DFM 3.8 release. Refer to Burt 327609 for more details, P.S. this issue is only specific to latency counters at volume level.