5

It should be simple, so it is just most likely my approach being totally off and someone will hopefully prod me into the right direction.

We have a Zabbix 2.0.3 server instance set up monitoring a bunch of different servers, but now we need to set it up to monitor and notify any alerts in regards to the SQL Server 2008 R2 Failover Active-Passive cluster.

Essentially, this is a 2 servers cluster, when only one of its nodes can be "active" at a given time, serving all SQL Server related requests, while the other server just "sleeps" and from the point of anyone logged on on that server - has all of the SQL Server related services in stopped state.

We have tried setting up Zabbix agents on both servers, using SQL Server 2005 templates (we could not find any 2008 specific ones and the 2005 ones always seemed to be working just fine for monitoring 2008 R2 instances) and configuring Zabbix server for both of the servers, but we end up having constant alerts for the server being currently the passive one in the cluster. We have been able to look up various methods of actually monitoring the failover, but we have not been able to find any guidance in regards to how to instruct Zabbix, that in this particular case, only one of the servers in the group is expected to be in the online state, while the other can be just discarded and should not raise any alerts.

I hope I made myself clear. Thanks for any guidance. I am out of ideas.

2 Answers 2

1

I do not know anything about SQL Server cluster, but there is probably a Service IP address associated with the active node that failover to the other node in case of HA event. Instead of monitoring SQL Server on both your nodes (hosts in zabbix), create a new host "SQL Server Active" with the Service IP address, and monitor SQL Server only on that host.

2
  • The thing is, that the Service IP is valid only on the private network, while the Zabbix server is located somewhere else. Setting up VPN between Zabbix and the private network could be a solution, but it adds another point of failure, which we would like to avoid. Also, this would not be enough, since we need to monitor on the OS level as well, e.g. the amount of free space on the data drive, and the iSCSI drive which holds data for the cluster is mounted on the active node server only. The best solution would be to have a group, where we could say: if one host is ok, disregard the others. Dec 4, 2012 at 8:49
  • My suggestion is to have 3 hosts: 1st node, 2nd node, and service. You could even have a 4th host to monitor your standby service. If both nodes are reachable, I do not think it is a big pain to make the service reachable as well. A well placed NAT could do the job here.
    – jfg956
    Dec 4, 2012 at 12:35
-1

I have 3 hosts in zabbix to monitor Clusters. 1 and 2 is NODES ofcourse. 3rd is Cluster. I monitor them checking on cluster what is his hostname currently so it introduces the active node hostname. I also found some way to check status through WMI. Just ask google :)

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .