Write latency numbers from hypervisor vs. storage

[RVil]

OS 3.8.1.0
Microsoft Hyper-V 2012r2 cluster

iSCSI connections

Periodic during Veeam backup, Veeam ONE reports too high read or write latency on different CSV volumes. A exampel is:

Cluster shared volume 2012 write latency

20-08-2017 23:29:05

"Disk/CSV2012: Write Latency" (272.0 Milliseconds) is above a defined threshold (80.0 Milliseconds)

But looking at infosight labs, i see the read/write latency on this specific volume during this time never has been above 3ms. How come Windows reports different latencies vs. the storage actual latency?

Robert

[dikrek · Accepted solution]

Hi Robert,

You can always call support for this but if the latency is very low on the array side and very high on the application, that usually signifies an issue either on the network or the host itself. For example, you may not have enough iSCSI sessions set up.

I can take a look if you give me your serial number, dimitris@nimblestorage.com

But first think about bumping the minimum iSCSI session number to 4. You've installed NCM on this host?

On the Windows machine, run “regedit” from the run window to bring up the Registry Editor. Navigate to HKEY_LOCAL_MACHINE > SOFTWARE > Nimble Storage > Connections > and find the MinConnectionsPerTarget entry. Make that a 4.

Thx

D

[RVil]

Thanks. NCM is installed, and i´ll start with raising the iSCSI connections to 4 and monitor the setup.

If this does not help, i´ll open a case.

Again thanks.