Datastore Latency Spike - VM metric only

[zcyd]

I have couple of VMs who's datastore performance metric show a astronomically large latency spike, 46000000000000ms. A user who was using the web console at the time reported poor console performance with repeatedly typed characters. From the host's performance metrics the datastore showed no spikes and looked normal. Of about 15 VMs that were checked during this time this is the only VM that showed the spike. 

vSphere 6.5. Ubuntu VM. ISCSI SSD storage.

Can this be explained?

[AishR]

These articles might be helpful to narrow down to the cause: VMware Knowledge Base and Troubleshooting Storage Performance in vSphere – Part 1 - The Basics - VMware vSphere Blog

[jstewx7]

I am having the same issue but with multiple VMs at sporadic time intervals. VMware reffered me to the manufacturer after finding HBA errors. Not sure what the issue is though at this point.

[christian_parki]

this KB may be applicable:

VMware Knowledge Base

Abnormal latency values are seen in ESXTOP and on the vCenter performance charts (2096171)

• esxtop displays abnormal DAVG/cmd and/or KAVG/cmd values.
• In some cases, negative values are reported for latency, which is theoretically not possible.
• In the vCenter Performance Chart you see unrealistic spikes in latency.
• During low I/O activity on a device, the READS/s and WRITES/s values are close to or equal to 0.