carmine wing
I have a 4 node cluster which consist of a FAS 2600 and a FAS2750. I want to decommission the FAS2600 and refresh with a FAS2800 but the challenges I currently have are :
Which other option will you advise?
frigid vessel
Mixed-version clusters are a thing: https://docs.netapp.com/us-en/ontap/upgrade/concept_mixed_version_requirements.html
We have used this a number of times in the past to upgrade older customer systems. Works quite well. But I would suggest to get the ONTAP versions as close as possible (i.e. update the cluster to the lastest 9.11.1 first)
carmine wing
We have used this a number of times in the past to upgrade older customer syste...
I have the 3 nodes in a cluster already but the challenge is how to relocate the aggregate from the 2600 to the 2800
acoustic kestrel
I have the 3 nodes in a cluster already but the challenge is how to relocate the...
You just move the volumes off of it.
frigid vessel
you cannot relocate the aggregate to a different HA pair. aggr relocation only works between nodes in a single HA pair. Maybe you wanted to do a headswap instead? You can do the headswap procedure while upgrading ONTAP at the same time, although you need to take care not to do anything wrong or you will (temporarily) lose access to your data
acoustic kestrel
would require an outage, but converting the 2600 into a shelf, connect to the 2800 and import the foreign aggr
fading crypt
Unless something changed This is not supported @acoustic kestrel with ONTAP. Used to be able with 7-mode.
Not saying it won’t work. Just not supported if something fits sideways
frigid vessel
yeah, I was looking for that because I also thought it was not supported but I couldn't find it. I know we had to do this once as well, when one controller died during a nondisruptive headswap... fun times, but we opted to do it with NetApp support together, just to be safe
I guess the cluster always finds the volumes by their UUIDs again, no matter where they are. and as long as the vldb ring is up it can update the location daatbase. But it was definitely easier in 7mode 🙂
fading crypt
That is significantly different @frigid vessel
Taking an aggregate from another cluster is very different than…a head failure in the same cluster.
frigid vessel
That is significantly different <@189749672921792512>
Taking an aggregate fro...
This is about taking the shelf from the same cluster, just a different HA pair? Or did I get this wrong?
fading crypt
I read it from another cluster
Converting the shelf(adding iom) then adding to another ha pair. Can be done but not supported
frigid vessel
but how will you connect the volumes to the correct SVMs then? all this info is lost with the old cluster? is there a way to manually re-attach the volumes to SVMs?
not that I suggest doing it, just curious 🙂
fading crypt
Probably part of the reason it is not supported
The aggregate is foreign. I believe once the aggr is part of the cluster, every volume has to be stitched into an svm. Maybe volume rehost? Maybe some manipulation of the cluster database? So reasons why it’s not supported
real nimbus
Dumb question, can you just move the shelves to the 2700 for now?
Or you could create the 2800s as a secondary cluster, SnapMirror to them from the 2600, then cut over the work, decomm the 2600s, then join the 2800s to the 2700s.
fading crypt
You can’t join active nodes to a cluster (meaning cluster setup is done). Only blank nodes can join a cluster
frigid vessel
You can’t join active nodes to a cluster (meaning cluster setup is done). Only b...
obviously 🙂 but as long as no data is on them it's relatively easy to wipe them again
fading crypt
Reread what he said…put data on and then join to 2700. No can do
frigid vessel
hm, yeah, not sure, first he said he has a 4-node cluster and then that it's a 3-node cluster. I was thinking that he just added the 3rd node so that it was easy to remove it again. But I probably didn't fully get what the status of the system is right now