HP ProLiant MP and Windows Failover Clusters

[Peter1135]

Hi!

We have an issue with failover cluster discovery. Besides the physical nodes it creates instances for all virtual
nodes of a cluster.

This means that if the cluster has two physical nodes and four virtual nodes, all monitored objects (Cooling, Memory, Processors and so on) are shown six times one time for each node.

Now, I created a group containing the windows computer objects of the virtual cluster nodes. Then I configured an override on the Proliant server discovery rule disabling discovery for that group.

Unfortunately this did not work - it still discovers all nodes.

The problem is that this generates duplicate alerts and notifications. For example if a NIC fails on one physical node you get 5 emails, one for the physical node and 4 for the virtual nodes.

Any suggestion is highly welcome.

Thanks,
Peter

[Stuart Cain]

If a system is identified as a Windows computer object and the system has HP Insight Management Agents or HP WBEM Providers running, the ProLiant sever MP discovery script discovers a HP Server instance. Can you try disabling the ProLiant Sever Discovery per windows computer object?

[Peter1135]

I tried to override the discovery per windows computer object instead of using a group containing all of them.
No success - all of them are still discovered and showing up in the HP Computers Group.

[Stuart Cain]

We want to confirm you are seeing all monitor objects and HP Server objects for nodes where discovery is disabled? If you disabled the ProLiant server discovery per windows computer object, we believe there are no objects discovered on these nodes.

We have seen where Operations Manager confuses the membership in the Computers view in HP Servers folder. If these cluster computer objects are associated to physical HP computer object (possibly by the Cluster management pack) then the Computers view displays all associated computer objects including the VMs.

[Peter1135]

Correct, I see all these objects regardless if I disable the discovery using a group containing the windows computer objects or doing it on a per object basis.

They are visible as group members of the HP Computers group (and all other groups depending on it) and in all views in the monitoring pane (Computers, Health Collection State, Server State).

[Peter1135]

Ok, now I tried to work around this issue. I configured overrides for all the Proliant SNMP monitors (Cooling, Temp Sensors etc.) to disabled them for the virtual cluster servers. Almost immediately those were shown as not monitored in the HP All Instance group. So I thought this will do the trick.

Today I realized: It does not. This morning a network link failed on one of our clusters and I got 4 alerts - one for the physical and three for the virtual nodes.

Next step: I will try to disable all the rules for the virtual cluster servers. Actually there are 92 Proliant SNMP rules. So it will take a while.

[Stuart Cain]

Yes, Alerts are generated from the Rules processing the SNMP Traps from the NT event log. The Monitors control State. Can you provide an update when you have modified the overrides.

[MarkE100]

You were close the first time.

Create a group containing the virtual cluster names.

Override the 'HP Proliant Server Discovery Rule' to disable it for that group

Fire up the SCOM Command shell and run 'Remove-DisabledMonitoringObject'. In a few mins the objects will be removed from the HP Group.

This useful command will remove any discovered objects that have object discoveries disabled against them. The way the HP discovery is written, it is a once only action and the objects will remain regardless of whether the discovery is disabled against them.

Cheers

Mark

[Peter1135]

Ok, I have configured overrides for all the alert generating rules targeting a group that contains the Windows Computer objects of the virtual cluster nodes. This stopped the duplicate alerts.

I will try the latest suggestion next week.

[Peter1135]

Finally I believe this issue is solved now. Sorry for the delay.

Here is what worked for me:

1. In the authoring space of the OpsMgr console create a group containing the Windows computer objects of all virtual cluster nodes.
2. Highlight "Object Discoveries". Enter "HP" in the look for field and click "Find Now".
3. Highlight the "HP Proliant Server Discovery Rule" and disable it using an override targeted to the group created in step 1.
4. Do the same for the "HP All Instance group member discovery". It is shown twice but you only need to do this on one object.
5. Open the SCOM Powershell and execute "Remove-DisabledMonitoringObject" without quotation marks.
6. In the monitoring space of the OpsMgr console expand HP Servers and select Computers to verify that all virtual cluster nodes are gone.

Make sure you add future virtual nodes to the group and if they have been discovered already execute step 5 again.