@Kaspars_Zibarts I know this is old, did you end up posting detailed instructions? I have a customer with a similar requirement - upgrade from R77 to R80 on new hardware with minimal downtime (FCU if possible). We also need a fast rollback plan that doesn't involve restoring from snapshots (ideally).
My initial thoughts are we have 2 options:
- Join the existing cluster
- As above, I'm hoping we can use cphacu but it doesn't sound promising due to different HW
- Will this even work if the interface names are different on the new hardware (vsx_util change_interface applies to the VS as a whole, not an instance on a member)?
- Rollback without snapshots could be a challenge if we have already upgraded the VSX object (see vsx_util downgrade caveats)
- Build a new cluster
- Replicate each VS on the new cluster (ports will either be shutdown or VLANs not allowed to avoid a form of split brain)
- Move cables (shutdown switchports / remove VLANs to old cluster, no shut switchports / add VLANs to new cluster
- Migrating one VS at a time is not really an option due to the heavy use of virtual routers & switches
The customer has roughly 75 VS across 2 clusters on the same MDS, option 2 sounds like it will have a better rollback path due to the independent clusters but requires all 75 virtual systems to be replicated on the new cluster (150 until the migration is complete, I think I will be getting very friendly with the VSX Provisioning Tool).
We have our SE looking for FCU-type options, fingers crossed.
cc @PhoneBoy ?