#Multiple backends in different Kubernetes clusters on one SVM

1 messages · Page 1 of 1 (latest)

clear lily
#

We have several Kubernetes clusters, each of them should have a nasbackend (ontap-nas) and a sanbackend (ontap-san-economy).
The different NAS backends will all have a different export policy which we specify in the driver.
The sanbackend should be connected via useCHAP=true.

My questions are:

Can different backend configurations coexist on one SVM ?
I think so.

Can different CHAP configurations coexist on one SVM ?
I am not sure about that.

Can the same CHAP configuration coexist on different Kubernetes clusters on one SVM ?
I'm not sure about that either, especially in the upgrade case. I assume that the existing LUNS are not affected, because otherwise you could never change the PW again.

fringe tartan
#

Yes, different backend configurations can coexist on one SVM, but it isn't recomended. If you do it, make sure you specify different storagePrefix values in each backend to ensure unique volume names.

As far as CHAP goes, I am doubtful that you can get it setup to work from multiple k8s clusters to a single SVM. I don't know that it has been tried as NetApp recommends one backend to one SVM.

clear lily
#

I was afraid of it.

But since we have very small and many workload clusters, many SVMs make little sense in our environment.
That would mean 2 SVMs per workload cluster (block/file).

I don't think we can manage that.

We are clear about the volumes due to the name label.

With iscsi, however, we have gone for the ontap-san-economy, so that the volume that is created does not have a storage_prefix but only the LUN name. Is there a possibility to name the “volume” name cleanly with ontap-san-economy? If that were the case, there would be no overlaps here. CHAP would then be out anyway and it would be igroups again.

foggy perch
#

CHAP has nothing to do with igroups or volumes though. it is just for (bidirectional) authentication. You can have multiple LUNs in the same volume mapped to different clusters. It's not a good idea (as has been mentioned by Scott) but you can definitely do it if you want to

clear lily
#

OK i understand

clear lily
#

Ok, so to make it clear again. There should be one SVM per workload Kubernetes cluster for each backend. In my case 1x NFS and 1xISCS. Chap is authentication, sorry for the misspelling.

I have tried it and nothing has happened. But again to be on the safe side.

If I delete the pvc in cluster01 and thus the lun. What happens to the volume trident_lun_pool_trident_TZQSPKQLTF. As far as I can see, nothing. Unfortunately it is not possible to keep the volume unique for each cluster.