Can't access CVMs in the cluster after change 1 GBE interfaces to br0

  • 16 December 2016
  • 2 replies

Badge +1
Hi last week we installed the first Nutanix cluster in Venezuela with 3 nodes, it was founded and as a best practice, the vSwitch br0 was "splited"into two vswitches by creating the br1 then the 10 GBE interfaces (eth2 and eth3) were assigned to the br0 and the 1 GBE interfaces (eth0 and eth1) were assigned to the br1, the administrator of the cluster decided to switch that assignment, i.e. to assign the 1 GBE interfaces to the br0 and the 10 GBE interfaces to the br1 by executing on the CVM's command line the command.

allssh manage_ovs --bridge_name br0 --bond_name bond0 --interface 1g update_uplinks

after the third node response to the command, the connection to the ssh sesion to the CVM was lost and now can't access the CVMs to revert the command neither ssh connection nor PRISM. In the nodes the 1 GBE interfaces leds are orange and red, only have access to the console by connecting a monitor and a keyboard to the physical nodes, my question is, does exist a way to revert the problem through this phisical console? what we can do?.. The cluster is under test now but it is intended to be in production next week. .. I open a case already.

Best regards

Best answer by Jon 16 December 2016, 18:40

View original

This topic has been closed for comments

2 replies

Userlevel 6
Badge +29
You mentioned you've got a case open already, thats the right direction to go.

Personally, unless you absolutely need this level of separation, I'd recommend just having one bond / bridge, with just the 10G connections.

The only time I've seen someone actually need it is when they needed to bring in a separate physical network, like a DMZ. Otherwise, just run all your VLANs through the 10G environment, and keep it simple.
Userlevel 2
Badge +17
The CVM command that creating the bond interface is just kind of script that automatically assign/create the underlying networking (OpenvSwitch) that you can reverse what you have done to be back to normal state before by using OpenvSwitch command

You can access to each physical node and execute the command below to verify your works:

- ovs-vsctl show (to show the openvswitch port table, you can also see the bond interface here br0-up from AOS 4.6)

- ovs-appctl bond/show [bond-name] (to show the properties of the bond interface)

If you are in mess now, I recommend deleting all bond interfaces as well as delete the br1 and start everything from scratch to control your works well