Too high datastore_totalreadlatency_average

[numeronove]

Sometimes, in particular under high IOPs load, those values are simply too high:

> SELECT esx,datastore,datastore_totalwritelatency_average,datastore_totalreadlatency_average
 FROM vsphere_virtualmachine WHERE datastore_totalreadlatency_average > 60000 LIMIT 1

name: vsphere_virtualmachine
time                 esx                     datastore     datastore_totalwritelatency_average datastore_totalreadlatency_average
----                 ---                     ---------     ----------------------------------- ----------------------------------
2018-03-24T16:20:02Z esxi3.dasecommercee.lan Eqldatastore1 116017258324865                     2049638230411510

That's a huge 3768,8 years.

I have encountered this behavior with VCSA version 6.5.0.13000, while in VCSA 6.0 it works fine.

It seems that the timestamp at the end of the measurement taken from VCSA, is going into the InfluxDB field for some reason. I have tried to look at the debug output, but don't know what I have to look for.

Can the new VCSA apply some unwanted formatting on the high IOPs value, or on the whole measurement row, under load?

[sofixa]

Hi,

It's funny because i've had the same issue, from times to times, with 6.5 ( iirc @exbane as well ), but hadn't actually gotten around to fixing it since i also use NetApp Harvest to get datastore-level stats from it, so not having correct VM-level data isn't the end of the world.

Have you had the possibility to check if the Monitor -> Performance tab in the vSphere Web Client reports similar values for latency ?

Cheers,
Adrian

[fullagame]

I have similar issues, https://imgur.com/a/uJ9z01g

Is this what you are referring to? Latency in the ESXi web client shows more expected values. :)

[codedmind]

Same issue vcenter 6.7.0.20100