Hello Team, is there any way to unjoin nodes without wiping out the configuration or erasing everything? for my understanding is not possible, however, I'm sure if maybe someone else did it?
#Unjoin nodes with out cleaning and erasing the existing config?
spice thistle
brave mirage
why would you want to do that?
if you remove a node, the cluster config on that node is useless anyway
burnt lintel
I have heard this ask before, typically just the ease of spliting up a cluster for mgmt reasons. but I don't think it's doable though.
brave mirage
but if you split it up you have two half-clusters with only half the data each?
burnt lintel
vol move what you want to nodes 1 and 2. vol move what you don't want on 1 and 2 to 3 and 4. and then split it.
brave mirage
aah. okay. I see, that actually makes some sense. And I think it could even be possible if there was a way to manually force epsilon onto a node after the cluster DBs are offline to bring them back online. But I don't think there's any way to do that, not even in diag mode 🤔 And it would probably leave tons of crud in the DBs that you couldn't get rid of and that comes back to bite you during the next ONTAP update or something
burnt lintel
yeah. not support/doable as far as I know.
brave mirage
supported? of course not 😂
the supported way: get some temporary shelves on nodes 1 and 2, move everything to nodes 1 and 2, unjoin nodes 3 and 4, set up a new cluster and use SVM migrate (or whatever other method you want) to move the data back. Then remove the temporary shelves once they're empty
we actually did this a couple of times. It all really hinges on the swing gear required. If it would require hundreds of TiB of SSDs we don't have that much in storage 🙂
given enough swing gear, you can solve basically every migration/reconfiguration challenge. nondisruptively 🙂
civic stag
given enough swing gear, you can solve basically every migration/reconfiguration...
..and if you have a nice NetApp rep. they will even lend or let you rent the swing gear 🙂 I know they have a list of hear on stock for this... (at least in Europe)... and we have used it a few times...
stone sandal
It happens that NetApp actually asks us if we have some gear lying around because apparently our stock is bigger than theirs. 😅
knotty talon
Unsupported ... but I've always wanted to try:
- Remove the HA pair from the cluster physically
- Treat each group of nodes as if the others have failed (there are processes to clean up a failed HA pair from a remaining cluster ... so just do it at both sides)
I've always wanted to try it.