Latency alerts for multi virtual disk vms

[breb]

I need some solution in case of having more virtual disks in different datastore tiers.

If I have as I have some vms with more disks in different datastore tiers, problem is which policy and alerts to apply.
P.e for virtual disk in "tier 1" latency of 20ms is warning, but if second or third disk are in tier 3, 20ms are OK.

In virtual disk metrics, there is only "aggregate of all instances" metric, I need separated alert for every virtual disk.
Is there some workaround to do it.

[KabirAli82]

There is no such option OOB as this could differ per VM, ea for VM1 it could be disk2 for VM2 it could be disk4 linked to that policy.

I guess the only option is to apply the lowest value you find acceptable to the VM (and thus to all virtual disks).

[breb]

Thanks Kabir,
came to the same conclusion, there is no solution.
We could set always the fastest drive (ssd) as a first drive, and end with the slowest one, but that means we will always get only false alerts. So configured should show: some of your drives is reaching configured threshold, but must not mean that any of the virtual disks is too slow.

[KabirAli82 · Accepted solution]

If thats the case maybe you should be able to pull it off...

I've created two Alert Definition;

First for the SSD:

 Mind the symptom for scsi0:0

 

The second for the HDD

 

Here it is scsi0:1

 

I think this should work across all your VMs as long as you always put the fastest on scsci0:0.

[breb]

Yes that could work.

But I have another problem, in Alert definition.
If I have 3-4 virtual disks, how can I create an alert for the p.e. third disk having in Metrics only 2  (scsi0:0 and scsi0:1) to select.

In Properties even less, only one.

[KabirAli82]

You're super close... try selecting the VM that has those extra virtual disks.

[breb]

Thanks, didn't checked that, that was a link. Anyway that is the solution, showing more disks.

Of course have another problem.

I have changed in Alert "Symptom definitions" some thresholds.
As a result half of the alerts using that changed symptoms are popping out triggered with old values and other with the new ones.
Updated for more than 1 hour. Is there some refresh / flush for alert definitons?

[KabirAli82]

Try putting the VMs in maintenance for 5mins or so.

Go to Administration - Inventory - Objects - <select your VMs> - this button:

[breb]

Thanks but no way, everything is in production.
I made the same brand new alert definitions, so it works for now.

[KabirAli82]

Glad to see its fixed.

Putting the VMs in maintenance in vROps is just stopping them being monitored for a brief period of time. Once they come out of maintenance vROps "refreshes" all sorts of data. I do this trick all the time when custom metrics/properties are not shown correctly.

Anyhows don't forget to mark the correct answer as the solution.