Does forcing a controller failover cause any loss of connectivity?

[frootmnig45]

Hi,

I have inherited a VMware system with a 4-node blade connected to a Nimble CS-215, I am now in the process of retiring the blade setup and replacing it with two rack mounted servers. The problem I have is that when the system was originally setup they did not put any switches between the Nimble and the blade interconnect bays so i have a problem connecting the new servers to the storage.

What I want to be able to do is re-route the connections from the non-active controller through the two new storage switches and then force a controller failover and then do the same with the now non-active controller, I do not have any experience with Nimble controllers and I want to be sure that forcing the failover will not cause any loss of connectivity to the current VM hosts.

[Nick_Dyer · Accepted solution]

Hello Simon,

As long as your external switch has the same subnets for Management & iSCSI with correct VLANs configured, as well as other networking specifics such as flow control, jumbo frames (if applicable) plus others such as spanning tree - then you should be ok. 

A controller failover will forcibly migrate the IP addresses of all array IPs from one controller to the other. This will also cause a re-ARP on the network, so you're looking at anywhere between 3-15 seconds worth of network "outage" whilst this occurs.

You will want to check your servers to ensure the correct Nimble Toolkits are installed for Windows, Linux and VMware to ensure that any applicable registry settings are applied, which will allow you to withstand this IP and re-ARP process.

If in doubt, give your local Nimble SE a call - and check in with Nimble Support to check your settings before you carry out this work.