Volume Monitoring for Clustered Nodes

Question

I'm trying to resolve an issue where I have an alert definition to alert when a volume becomes unavailable. This works well for server nodes that are not part of a cluster. However, for clusters I find that a volume alerts as unavailable when the the other node becomes active. Tjese are Windows clusters. What am I doing wrong?

adam.beedell · Answer

Aah, ok I understand now, I think in SQL clustering you can query that with a SQL command, so you could get your data from an application monitor with a "SQL user experience monitor" component in it. If your cluster is something else I'm not sure how you'd poll it off the top of my head.

If you can poll it:

* Make an application monitor that polls it with SQL or powershell or something
* Set an alert that checks for the result of that application monitor, and writes to a node custom property
* Filter out passive nodes in your original alert based on the custom property

If you cant poll it but you know which ones could be passive you could:

* Stick their servernames in a spreadsheet, give em a custom property, upload the data and filter them all out forever

Or

* Stick their servernames in a spreadsheet, give em a custom property, upload the data
* Duplicate your current alert, filter it to ONLY those with the custom property
* If the alert would trigger, instead of sending an email, write to a 2nd custom property
* On reset clear the 2nd custom property
* edit your original alert, filter out anything with the 2nd custom property
* Set the duplicate alert to trigger faster than your original alert

That's a bit jank but I think it'd work in either case

jameslindsay · Answer

Correct. I need some metric or value that indicates that it is passive that I can use in the alert to suppress the trigger. I can't really be chasing which is active and which is passive with a custom property unless that value is set automatically when the switch occurs.I would have hoped I could pick that up with Microsoft Windows Server 2012-2012 R2 Failover Cluster. But, that doesn't even seem to know who the members are much less who is active and who is passive

adam.beedell · Answer

That sounds like the alert's doing what it's supposed to be doing - When the node in the cluster is passive it doesn't have access to the drives, right?

Perhaps you filter the clusters out using a custom property or name format?