cancel
Showing results for 
Search instead for 
Did you mean: 
Create Post
Level 10

How to properly monitor a SQL cluster with Orion NPM 10.1.2

I want to monitor a SQL database that runs on a cluster, but I have been reading Thwack posts about Orion "losing" track of the physical resources in the event of a failover.

Can anyone help me with a step-by-step of how to add the cluster to NPM and how to monitor the resources of the cluster and alert if there is a failover? Would I use APM for any of this?

Thank you so much for any help.

0 Kudos
12 Replies
Level 8

I just thought I would chip in and offer this PowerShell script to monitor MS Cluster resources. You can run it against the cluster to see if any resources are down, against a virtual server to see if any vs is not running a resource, or against a physical node, which would show which resources are running on a physical node.

 

$cluster = Get-WmiObject -ComputerName '${ip}' -credential '${CREDENTIAL}' -Namespace "root\mscluster" -Class MSCluster_Resource

$downlist = @{}
foreach ($c in $cluster) {
if ($($c.state) -ne 2) {
$downlist[$c.name] = $($c.state)
}
}

write-host "Message: Unavailable resouces: $([string]::join(', ', $downlist.Keys))"
write-host "Statistic: $($downlist.Count)"

When I run this powershell I get the the follow Message: Unavailable resouces: System.Collections.Hashtable+KeyCollection

0 Kudos
Level 14

You will probably be able to get some of what you want by monitoring each server in the cluster with both APM and NPM. Basically, you will want to do the following:

  1. Add each server, individually, as a node in NPM.
  2. Add the cluster IP as a node.
  3. Add SQL Server application monitors on each cluster server.
  4. Configure alerts in NPM to notify if/when any clustered server goes down.

Unfortuantely, I don't think we can currently provide any visibility into which clustered server is active at any given point, but I've marked this thread for PM, so it can be taken down as a feature request.
Thanks,



You will probably be able to get some of what you want by monitoring each server in the cluster with both APM and NPM. Basically, you will want to do the following:

 

  1. Add each server, individually, as a node in NPM.
  2. Add the cluster IP as a node.
  3. Add SQL Server application monitors on each cluster server.  on the cluster IP
  4. Configure alerts in NPM to notify if/when any clustered server goes down.
Unfortuantely, I don't think we can currently provide any visibility into which clustered server is active at any given point, but I've marked this thread for PM, so it can be taken down as a feature request.
Thanks,

 



 

There, I fixed it.  This is what we are doing and it works.  No need to know what node is active.  We also have servers with multiple instances with some running on each node, so we always monitor the instance IP for SQL services, or any other cluster resources.

0 Kudos

Well, the method is correct but there are two additional points...

Yes you can get visability into which node "owns" the cluster resource.  The "system Name" will change depending on the where the resource is and if you use HP server the Cluster monitoring built into the HP Management agents support the oids that give you status information.

We have an alert that fires when a cluster fails over based on the system name.

JB

Hi JB-

Can you give some information on how you are getting ClusterResourceOwnerNode? Is this done through a Universal Device Poller? What MIB ID do you use?

Thanks,

Keith

0 Kudos

I would like to know how to set this up as well.

0 Kudos

They are available in the HP Managements Agents (CPQCLUSTER-MIB table) 1.3.6.1.4.1.232.15.2.4.1

JB

0 Kudos

Has there been any update on this?  I have recently moved to a new employer and I am tasked with rolling out Solarwinds to monitor their network.  There are several clustered systems here that I am curious as to the best method to utilize SolarWinds to monitor them will be.

0 Kudos

In addition the this you want to make sure that you are only monitoring 'Direct Attach' volumes on the physical servers and 'SAN' volumes on the cluster IP.  In this way you maintain continuous view of the clustered volumes (the ones that get passed back and forth between the physical servers during failovers).

0 Kudos

Is it possible that custom pollers could be used to track the assignment of the clustered volumes at any point in time?

0 Kudos
Level 10

No one?

0 Kudos