Anyone monitoring MSSQL on a Windows cluster?

[tchjts1]

We have a Windows cluster setup - two physical nodes that create a virtual node. Zabbix is installed on both physical nodes. There are two shared/clustered drives, Q and S.

The MSSQL DB resides on one of the clustered drives. I am using perfcounters to get the metrics I want, but when there is a failover from one node to the other, Zabbix loses the metrics and they go into "Not supported".

Does anyone have a solution to this? I think I am going to test installing Zabbix to the virtual node on the same drive where the DB resides and see if that works, but if someone already has the answer to this issue... I'm listening.

Thanks -
Jeff

[bagni]

Hi,
if you have a failover cluster the istance must be have an VIP (virtual ip), so I usually monitor the physical node with active check (ram, swap, disks by autodiscovery, etc.) and with passive check the SQL status and performance counter by VIP.

If the istance switch to the other node the VIP will be the same, so the monitoring don't shoot the trigger.

[tchjts1]

Thanks bagni, I will take a look at that. I have tried both passive and active agent checks to the physical node IP's, which didn't make a difference. I'll try to the virtual IP.

However, restarting the agent fixed the issue, which is perplexing.

[bagni]

Use the disks LLD on physical nodes to avoid the restart of agents.

[tchjts1]

Use the disks LLD on physical nodes to avoid the restart of agents.

Not sure what you mean. I use LLD on all of my servers. How does that affect not having to restart the agent? The agent is not losing all MSSQL metrics. It appears to only lose 5 or 6 (out of many) when there is a failover. I need to get with our DB team and understand the DB config a little more.

By the way, are you doing any alerting on to notify you when there is a failover?
If so, what trigger are you using?