1

I'm trying to monitor SQL Server instances on a 2 node active/passive windows failover cluster with Enterprise Manager.

Oracle gives the following steps:

-Install local agents on both nodes.

-Install agent on shared storage for the virtual hostname of the cluster on ONE node.

This seems to be somewhat incomplete, especially the last part.

Agent installation creates a windows service. Creating a failover resource would require said service to run on a shared storage and to be defined on both cluster nodes.

So I'd expect the service for the virtual cluster is either supposed to replace the two local agent services or it needs to be copied to the second host? Which is the better option or am I completely wrong? (My replacement theory is based on my puzzlement that a local agent neede to be installed on each node in the first place, why would that otherwise be necessary?)

I expect to run into additional problems when I create a cluster role from the agent service, since I can only do that by giving the role its own IP. However, the agent is registered with the OMS host on installation. Will I need to also fix the data in the repository for the new IP adress in an additional step, or will the combination of (virtual cluster adress):(agent port) automatically point to the role's new IP?

Sascha Rambeaud
  • 121
  • 4

1 Answers1

1

So, it turns out most of the information on this out there is outdated.

As of version 13, the process is:

  1. Install agent locally, once on each of the nodes.
  2. Using the emctl command from the locally installed agent, deploy the agent, using a volume that is either shared or can failover, on each node.
  3. Create a cluster role for a "generic service" using the service created in step 2 OR add a ressource of type "generic service" to an existing cluster role
  4. Repeat steps 2+3 for every SQL server instance on the cluster, that has its own host name if you wish to discover those SQL server instances
Sascha Rambeaud
  • 121
  • 4