Monitor applications clustered using Microsoft Cluster Server Technology

[sourashis]

Hi,
If I have a cluster being managed using Microsoft Cluster Server (MSCS) how do I monitor the cluster? Does Hyperic provide any support for MSCS clustered environment or do I need to install agents on each of the machines in the cluster individually and then define my own group/cluster to monitor them? I will appreciated any help in this regard.

Regards
Sourashis

[BradFelmey]

I believe an agent per. That's how we have our four-headed Exchange setup done.

[petervdb_hyperi]

If a services switches from node and for example the P drive disappears on node A and gets active on node B, how does Hyperic reacts on this?
I would assume that an alarm will be generated because on nod A the P drive is not available anymore. But if the P drive is available on node B the alarm should be cleared.
Can Hyperic handle these situations?

[BradFelmey]

The short answer (insofar as my experience with this goes) is no.

HQ doesn't really seem to be aware of MSCS at all. If node 1 goes offline and node 2 picks up, you get an alert that node 1 is down, and the dashboard reflects that the IP set changed for node 2. This is something that has to be "imported".

[BradFelmey]

With that said, I don't know if the policy-based alerting in the Enterprise Edition would allow for something like this. We simply haven't tried tying one alert to another (different) platform.

[petervdb_hyperi]

OK, I will start installing the agents on the 2 cluster nodes and see how far I can go ...

[SLTB]

If you run seperate agents on both nodes you miss the point of cluster.
You will constantly see services which are unavailable, just because they are on the passive node.

I don't have a cluster to test, but I assume the following can be tested:

1. Install hyperic agent on both nodes
2. Create a managed cluster resource of generic service for the agent

The agent will run under the cluster virtual IP.
Each failover will cause the agent to stop on one node and start on the other node. Then the agent will perform autodiscovery and monitoring localy on the relevant node. If the nodes are a reflection of oneanother this should help to monitor it as a cluster.

Message was edited by: SLTB

[BradFelmey]

Can you expound upon #2?

[SLTB]

This is something internal in ms-cluster. You can define a generic service (windows service) which depends on the network name of the cluster (and not of a single node).
You configure the command line that the generic service will run. The service can run only on one node at a time (the node in which the specific group of cluster resources is active). If you can ensure that the same command line in each node will run the agent with the appropriate plugins installed on it, ms-cluster will give you the active/passive clustering. When performing fail over of the group the ms-cluster will shutdown the agent on one node and run it on the other.

In theory this should work, but I don't have a cluster to test it.

[BradFelmey]

We do have our Exchange cluster set up like this. I thought you were talking about an HQ cluster service, and I was looking all over for it. 😉

[SLTB]

If this is working and supported you should add it to documentation.
I believe others would want it too.

[satyabrata]

Hi,
I need to know more about exchange cluster monitoring through hyperic.
Please let me know the details.


Thanks,
Satyabrata