Switch Stack replacement - What to do for a 3 nodes Ceph cluster

[TwiX]

Hi,

I plan to replace a stack of 2xswitchs. A 3 nodes cluster (with dedicated lacp interfaces for vmbr0 and ceph) is connected to this stack. I guess I have to shutdown the entire cluster for doing this. What are your recommendation for doing this properly ? (Stop all vms, then what just shutdown all 3 nodes ?)

Thanks in advanced !

 

[Alwin]

Depends on how you connected your cluster to the stack. But if you can setup the new switches on the side, then you could add a trunk between the old and new one (including VLAN, etc...). It would then be possible to switch cables one by one during runtime (VMs migrated away).

 

[TwiX]

Hi, good idea but I dont know what could happen with lacp (remove a cable from a lacp LAG to another port which is not a member of the old LAG). If you know what I mean

 

[Alwin]

I would go and configure the LACP on the new switch and change all cables at once, as the switches are connected by trunk. You set noout for ceph, so it isn't going to recover.