08-18-2012 01:26 PM
I've been reading the manual regarding bonding the NICs on the P4500 system that we have. It states that you should bond the NICs prior to adding the nodes to a management group. It seems the team I work with were in a rush to get storage online and they added the nodes to a management group prior to bonding the NICs.
My question is am I still able to bond the NICs on the nodes even though they are in a management group? Will I expect to have any problems if I choose to bond them while they are in a management group? Can anyone recommend a course of action in this circumstance?
Solved! Go to Solution.
08-19-2012 12:24 AM
It depends on your network setup. If you plan to use ALB or active-passive bond, then make sure that both NICs are connected to correct vlan and just create bond via CMC.
If you go with LACP bond, then things are a little more complicated. First, you need to create on switch LACP trunk with one inactive node's NIC and place this trunk in correct vlan. Then create LACP trunk from CMC. Node will negotiate with switch via one NIC. Then on switch add link to second NIC (which is active now) to LACP trunk.
08-20-2012 05:46 AM
Thanks for the reply. I am planning on creating an ALB bond. I just wasn't sure if I should expect any problems. The manual states on pg. 97 that if you create the bond after the nodes have been joined to a management group, and it doesn't work correctly you might:
- Lose the storage system from the network
- Lose quorum in the management group for a while
I was curious if anyone else has been in my situation and what their experiences are....
08-20-2012 05:53 AM
as with any network change you are making to a live system, you will probably have the node experiance some downtime. since you have two nodes and are running network raid 10, this should not be an issue as long as you have the management group setup correctly... you do have a FOM right? If so, you should be totally fine.
I don't know if it really makes a difference, but I like to stop the manager on a node I"m going to reboot before I reboot it just because I feel it makes the failover quicker when the node that reboots isn't the coordinating master.
08-20-2012 06:20 AM
We plan on bringing all servers that are connected to the nodes down just to be thorough. We do have a FOM in place.
I will take your advise and stop the manager prior to rebooting the first node I make the change on. We have two nodes in a cluster, so likely I will bond one node, reboot, wait for it to come back up then do the same to the other.....
08-20-2012 10:43 AM
Network bonds can be deleted/created online. If I remember right, during bond creation just manager process disconnects (and instantly reconnects) from quorum. There is no need to reboot nodes for this procedure.
08-25-2012 02:42 AM
Disconnect from current network. Make sure it is going to stay up on the single node. (follow the rules, have a FOM etc)
Attach a screen and keyboard or fire up the iLO. Attach both ports to a (dumb) switch and a laptop.
Apply an IP address to laptop NIC in same range as P4000,
Fire up the management console, make your changes. Expect it to not work nicely as it often does not. (this is why we disconnected it from the network - this way it should not stuff up anything)
Work your way through until things are how you want them.
Reconnect, allow things to rebuild/sync in their own time.
ALB is easy, should work with most switching infrastructures - good choice :)
Lots of people are tempted to try and make LACP bonds work over seperate (unstacked/not real stacking) switches. Does not work.
08-29-2012 11:23 AM
What we ended up doing was:
Shut down all servers that had a connection to the cluster
Created the bond on the first node in the cluster
Created the bond on the second node in the cluster
Brought the servers that had a connection to the cluster back online
Everything went smoothly!
Thank you all for your input.