Skip to main content

Hello All,

I have a question regarding the behavior of an SQL cluster.

 

Scenario:

This cluster is made up of 2 VMs with shared disks. There is also a witness client to this cluster. The witness client was rebooted unexpectedly, and so I’m assuming Commvault determined that the cluster was lost, since a backup set and instance were automatically created on one of the client nodes. What brought our attention to this was that we received an alert that this subclient could not be backed up since there was no Storage Policy associated, and permission not configured, which was to be expected.

My questions:

  • Is this normal behavior? (I believe that it is?). If so, why would Commvault do this given it can’t be backed up anyway with no storage policy associated, etc...?
  • Is there an option to disable this behavior?
  • Am I correct in saying that by the witness client being rebooted, the system assumed the cluster was lost? Also, this is not the same as a failover, yes?
  • If an actual failover occurs, would Commvault be able to continue backing up the SQL DB without intervention?

 

Thanks for any help or relevant docs on this topic

Cheers!

Hi Lani,

 

Whilst autodiscovery of SQL instances is the default behaviour here, I’ve never seen a clustered instance show up under a physical node automatically. I have seen people add them manually though. It’s interesting that backup had been scheduled for the physical node, however this could just be that the node is a member of a group that has a schedule already assigned I believe. 

 

If you want us to dig deeper into that, it would be great to have a ticket raised with support for further analysis.

 

In relation to the questions:

 

Is this normal behavior? (I believe that it is?). If so, why would Commvault do this given it can’t be backed up anyway with no storage policy associated, etc...?

 

I’ve never seen this behaviour in the past. I wouldn’t expect a failover cluster instance to be automatically discovered under a physical node. When instances are automatically discovered in Commvault the subclient is created without any storage policy associated. The storage policy should be manually associated prior to backup being attempted.

 

Is there an option to disable this behavior?

 

SQL instance autodiscovery can be disabled by unchecking the autodiscover option on the agent properties.

https://documentation.commvault.com/2023e/expert/enabling_automatic_discovery_of_sql_instances.html

 

Am I correct in saying that by the witness client being rebooted, the system assumed the cluster was lost? Also, this is not the same as a failover, yes?

 

I am not sure of what occurs to a Windows Failover Cluster when the witness server is rebooted, however I don’t believe that a failover would ensue. To my knowledge the witness server is there to provide quorum to the cluster when an election is required. I am not sure how this would have impacted Commvault at all in this scenario, as we would only have agents configured on the cluster members.


If an actual failover occurs, would Commvault be able to continue backing up the SQL DB without intervention?

 

Yes, the cluster plug-in that is deployed as part of pseudoclient configuration will come online on the current active server and update the Commserve with the current active server. This will send all future requests to the current active server. 

 

Cheers,

David

Hi David,

 

Thanks very much for your answer, and all the details you provided. Very much appreciated!

As you suggest, I will raise a support ticket to have this investigated.

 

Thanks again!

Lani

Reply


Terms & ConditionsCookie settings