Disk latency

[binrad]

What does Disk latency mean with respect to virtual disks?

[Dynamax]

how are you viewing this? esxtop? if so you have 3 latency columns. DAVG = device latency and it the latency of requests from HBA to storage, KAVG = kernel latency, GAVG = guest latency which is DAVG and KAVG added to together.

[binrad]

The Vsphere client shows a triggered alarm stating the disk latency is beyond the default threshold for the datastore where the VM resides.

thanks,

binu

[Dynamax]

it'll be the latency of requests made to the backend storage. Whats your set up? is this the only vm on this data store? whats the backend storage of this data store?

[binrad]

Set up is a FC (EMC-CX4) of 100GB VMFS data store and it has 9 VMs.

Path selection policy is set to "fixed " . Out of 9 VMs, one is configured as a file server and the users are not able to access the files located on file server VM.

[Dynamax]

not sure that would be a latency issue. Whats the RAID group config i.e RAID type, no. of disks, how many luns allocated from that RAID group. For disk latency you may not have enough spindles allocated to that RAId group

Can you get other connectivity to this VM like RDP, ping etc.

Also im pretty sure that the recommended setup for CX's is MRU. Shouldnt be an issue for this though

[TimOudin]

MP selection caught my eye as well, though highly unlikely that it's an issue at hand here. Just insure that you've read and are in compliance with the clariion integration doc if you are going to use a FIXED policy and set the failovermode appropriately!

http://www.emc.com/collateral/hardware/white-papers/h1416-emc-clariion-intgtn-vmware-wp.pdf

Tim Oudin

[TimOudin]

Do you have Navisphere Analyzer? Can't hurt to take a look there for high response times on your LUN... Can't tell you what metric vCenter is reporting, sorry.

Tim Oudin