vsa managers and failover

[danletkeman]

Hello,

I have a three node vsa cluster with one vsa per physical box.  When I take one of the physical boxes offline, I lost connection to the san.  My cluster shows that I have 3 of 4 managers running, 3 managers, 1 virtual managers.

Is there something else I need to do to make this redundant?

Thanks,

Dan. 

[ajamil786]

Three VSA (managers) are redundant. You don't need a virtual manager. 4 manager configuration is not recommrnded. You should either have 3 or 5 managers under normal setup. With that said you still should not loose connectivity to the cluster. What errors atre you seeing?

[danletkeman]

Looks like what happened is:

The quorum communication status for manager 'DC1-VSA2' in management group 'DC1-VSA-1' is 'Down'.  Manager 'DC1-VSA2' cannot communicate with managers 'DC1-VSA1, Virtual Manager'.

We took down the DC1-VSA1 host, and the DC1-VSA2 host could not communication with DC1-VSA1 virtual manager.  So I have deleted the virtual manager and I will try again.

[ajamil786]

To me ity sounds like you had already started the virtual manager and it was started on 'DC1-VSA1'. When 'DC1-VSA1' went down so did the virtual manager. You have to be careful with the virtual manager. It is merely useful when you have an even number of notes specifically 2 nodes and one of them went down. You can in this case start the virtual manager on the existing node to regain quorum until your physical node comes back up. In a two node environment you can add the virtual manager to the management group but do NOT start it.

[RonsDavis]

You definitely want to remove the virtual manager from the management group. If it, and a node are down you have lost quorum. With three nodes each should be a manager, and as long as you don't loose two at once, you will keep quorum, and all will be fine. Oh course, if you have any volumes set as RAID 0, they will go down if you lose a node.