cancel
Showing results for 
Search instead for 
Did you mean: 
Create Post
Level 9

Monitoring multiple AGs on the same instance

Hello

We have an instance that has 2 availability groups.  If we setup monitoring on each AG there is overlap for data collection. In DPA, we see the activity for ALL databases on the server, not filtered by the AGs. This means that we have 2x the data collection that we need.  If we remove one of the AGs from monitoring, then we won't have any alerts for AG related issues.

What's the best way to set this up?

Thanks!

0 Kudos
4 Replies
Level 9

Hi

The general recommendation from Solarwinds is:

Register each SQL Server instance in the cluster

If there are multiple AGs in the cluster, this option is recommended because it ensures that DPA does not monitor the same instance more than once. DPA monitors all activity on each instance, including primary and secondary AG activity.

With this option, DPA does not follow AGs when they fail over. Monitoring all instances in the cluster ensures that you see all activity when AG failovers occur.

-> Registration and licensing options for clustered environments - SolarWinds Worldwide, LLC. Help and ...

Another useful article regarding AG: About monitoring SQL Server Availability Groups - SolarWinds Worldwide, LLC. Help and Support

regards

0 Kudos
Level 9

That is true but then the charts (for performance history) have huge gaps when a failover occurs.  also, wouldn't that screw up the baselines?

0 Kudos
Level 9

jsql  wrote:

"That is true but then the charts (for performance history) have huge gaps when a failover occurs..."

That's correct.

The other question is: How often do you have a failover situation? Is it a part of the application logic or is it only happen in an emergency case and then you can life with the gap?

jsql  wrote:

"... wouldn't that screw up the baselines? "

A question for the experts. Probably this is handled with the new feature of 12.1. "Anomaly detection" ?

0 Kudos
Level 9

it's not often that we have a failover, i'd say one every couple of months, usually a network issue. but once we have a failover, we don't failback, there are some applications that are sensitive even though they are "AO compatible".  we usually live with the failover until the next windows/sql patching and them sort them out back to the original cluster node. could be a couple months. in these cases, merging the history between the 2 nodes is not very clear for management reporting.

0 Kudos