nlc2013
Contributor
Contributor

Lost connection to hosts when disconnecting 1 of 2 enclosure switches?????

Jump to solution

Hi

Hoping someone can help point me in the right direction for this problem.

We have three IBM HS23 hosts setup in our cluster, these hosts all sit in the same enclosure and said enclosure has 2 Cisco 3110 switches in the back. Both of these switches are configured with etherchannel trunks to our core switch stack and carry 10+ VLANs over the trunk. These VLANs are configured as port groups under a single standard switch on each host and the guest VMs are dropped into the relevant VLAN and can communicate across the network.

The problem we have is that yesterday I temporarily unplugged the etherchannel cables on the second Cisco 3110 in the back of the enclosure and we lost connection to 2 of the 3 hosts from the core, plugged the cables back in and connectivity was restored. I then unplugged the cables from the other Cisco 3110 and we lost connectivity to the remaining host.

I was under the impression that as both available vmnic on each host are active adapters for each port group on the standard switch that we had resilience in the system whereby we could lose one Cisco 3110 and maintain connectivity.

This isn't the case but looking at the config I cannot see why.

Has anyone got any advice on where I need to be looking or experienced something similar?

Thanks

0 Kudos
1 Solution

Accepted Solutions
a_p_
Leadership
Leadership

The ESXi hosts are connected to the internal switch ports of the two Cisco switches, and unless you have configured Link State Tracking on the Cisco switches (i.e. if the uplink ports are down, the downlink ports will be shut down), the ESXi hosts will still see their internal Cisco ports online, and therefore will not fail over to the other switch.

Andé

View solution in original post

0 Kudos
3 Replies
a_p_
Leadership
Leadership

The ESXi hosts are connected to the internal switch ports of the two Cisco switches, and unless you have configured Link State Tracking on the Cisco switches (i.e. if the uplink ports are down, the downlink ports will be shut down), the ESXi hosts will still see their internal Cisco ports online, and therefore will not fail over to the other switch.

Andé

View solution in original post

0 Kudos
nlc2013
Contributor
Contributor

Thanks for your reply

I've been doing some investigation and I'd come to that as a feasible conclusion but thanks for confirming it.

Would enabling beacon probing in the switch properties help with this? I've read that it isn't suitable for only 2 physical NICs (which is what we have) but if those NICs connect to two different switches in the enclosure can they actively monitor the links?

Thanks

0 Kudos
tedg_vCrumbs
Enthusiast
Enthusiast

Beacon probing is not recommended with etherchannel and does require 3 NICs to properly identify a failure.

------ tedg Don't forget to mark posts as helpful or correct if they deserve it!