We have a well established XCP-ng pool with three hosts.
These have a basic setup of:
eth0 = SAN Route1
eth1 = SAN Route2
eth2 = Management
eth3 = VM Data (Bond0)
eth4 = VM Data (Bond0)
eth5 = Unused
eth6 = Only two hosts have (Unused)
eth7 = Only two hosts have (Unused)
vlans = linked to bond0 (eth3+4)
My questions is how to I go about this upgrade. A) Add SFP card and use PIF renaming to change the order of the NICS. I read this should be done before joining the pool but this is already a production pool. if it is possible to do at this point, i guess i would down all NICS on all hosts and rename as required, then bring it back online?
B) rebuild one host from fresh and reorder the NICS on build. Join the pool and become the master, then rebuild the other hosts. - this method assumes the hosts won’t mind some hosts being copper and another being SFP for the same ethX advisable?
C) build a new pool and migrate settings and VMs over to new pool. I guess i could export all VMs metadata, attach storage and then import metadata to new pool.
I prefer B but would like to shortcut with actions from A, maybe not a full rebuild?
You can simply reset the networking, do another reboot, and it should come back up fine. I did this with my test system a while ago and after some initial confusion, it popped back to life and the other 2 hosts were simple to put the cards in and get working. I think the process was hands on the server, delete old NIC, shutdown, install new NIC, power up and configure new network with old IP info. Then it didn’t take hold until I did another reboot. It was a while ago and I went from 1gb to 10gb. I did fight with the first one a lot, but the other two just took a little time. I think the fighting was my own lack of experience.
Keep in mind, I only had a single connection and single vlan to worry about. If you chose the rebuild route (plan B), don’t forget to reassign the pool master away from the unit you are working on before it gets shutdown, it’s a real hassle if you forget, especially if the server is slow to boot.
As far as the copper vs faster issue, I don’t think that will be much of a problem unless the hosts need the faster connection. And my guess is that you are going from slower to faster so they won’t care.
I’d also put up a stand alone XO for the time while you are doing this work, just some old workstation you have sitting around with debian and XO from sources. Or a Windows box with XCP-NG Center.