We need to migrate data from an older FAS2620 to a new F50, we can have a bit of down time if required (also this is a single node FAS2620)...
I was thinking of adding the new F50 into the FAS2620 cluster using some "older" NetApp cluster switches (C1610)... but FAS2620 is stuck on 9.11 which I don't think F50 supports? Is it possible to add nodes with a later ONTAP version to a cluster? (my guess is not...)
What are the alternatives? I would rather not SnapMirror all this, because we have existing snapmirror relations which will then have to be broken and resynced etc...
Options I can think of is so "somehow" attach the existing disks to the F50 (which already will have a full DS460C) and the FAS2720 has 8 shelfs attached (96 disks) and of cause disks in the controller as well ( which we can move to an empty shelf ).
But I am not sure if we can just do a "headswap" from a single node to a HA setup?
We have some spare FAS2720 which we could use, and migrate to them first, which will then allow us to upgrade ONTAP and add the F50 to the cluster...
But it is as easy as to just headswap the controller to a FAS2720, and then add the other node (there are empty root-partitions) ?
I guess using the C1610 switches will work, but I also guess it's not supported? Do we need a PVR for a thing like this?
#Migrating data from FAS2620 to F50...
sly venture
I can already see one issue... the F50 has 40/100Gb cluster/HA ports... and the C1610 only have 10G ports.. can this be "fixed" with a cable-splitter? 40G to 4 x 10G maybe? And I guess the question about this will be "blessed" by NetApp goes out the window right there? But is there a supported switch for a case like this?
glossy hill
mixed-version clusters are supported, but only for a limited version-difference, and yours (9.11 to 9.16) is sadly just barely not supported (9.11 to 9.15 would be supported, as would be 9.12 to 9.16)
if you have enough spare slices (or disks) for a second root aggregate, an offline head swap will work (since the FAS50's minimum root aggregate size is smaller than the FAS2620's)
- shut down the old controller
- convert it to a shelf and connect it to the new nodes
- boot the first node into maintenance mode
- reassign the disks from the old controller to this node
- boot the node, answering "y" when asked to make sure the system ID change was intentional
- boot the second node and let it auto-assign enough root spares (or do the disk assignment manually)
- join the second node to the cluster
afterwards you have to fix the network ports/ifgrps/etc. since the port names have all changed, of course (and delete the old no longer existing ports with network port delete)
your second option would be:
- headswap the FAS26xx to the FAS27xx (offline only, basically same procedure as above)
- upgrade the 27xx to ONTAP 9.16 (or whatever your FAS50 will run)
- add the second node on the FAS27xx (create root aggr, join the node to the cluster)
- do a headswap 27xx -> FAS50 (HA system head swap, so this should be nondisruptive)
your third option (which only works if your FAS50 comes with new disks, and if you can find a switch that has 40g and 10g ports since you cannot use breakout cables on the FAS systems):
- headswap the single 26xx to the 27xx (disruptive, as above)
- upgrade the 27xx to 9.15 or 9.16
- connect it to the switch, create cluster ports
- connect the FAS50 to the same switch (make sure to use the same config as the RCF file, i.e. VLANs etc)
- cluster join the two nodes, building a 3-node cluster
- vol move everything over to the FAS50
- decomission the single node system
This is a non-supported operation but it should work if you carefully check every step
if you had a second 26xx controller, you could even make the headswap completely nondisruptive with the help of the 27xx controllers
so, depending on the timeframe/SLA requirements, and whether you can source a FAS2620 controller, I would either do the direct offline headswap with downtime, or try and go with the full nondisruptive one which takes a bit longer
sly venture
if you have enough spare slices (or disks) for a second root aggregate, an offli...
OK, and booting up the F50 with an root aggregate from the FAS2620 (with ONTAP 9.11) on it will "work"? Will it just update it to 9.16 or what ever the F50 has installed? But after this I will then have a task of moving the root aggregate to the DS460 shelf, but I think there is a procedure for that task... I think this would be the better option because we can skip the FAS2720... (and yes some downtime will be OK)...
glossy hill
hm, yeah, you're right I haven't tested such a large RAID level discrepancy myself...
sly venture
...so maybe do a swap to FAS2720 first would be the safer bet here 🙂