we're transforming a 2554HA 7Mode in a 9.3P6 clustered couple of nodes.
We've disabled the HA in the original 7 Mode, assigned the disks to the second node and we've performed with success a cDOT setup of a two nodes cluster where, temporarely, there's a ghost node, the second one, that will come once the migration have been completed (it will be the node that now run 7Mode)
So, we've a running two cluster with one node waiting for the second.
We need now to migrate via FCP a VMware datastore content from the LUNs on the remaining 7 Mode controller to the new one.
Problem. FC SVM created, FC LIFs ok, a couple of, LUN created, igroup created...NO MAP. There's no way to associate the igroup because there's no FC LIFs on the expected second node.
We've tried to disable cluster, SFO and so on but the issue remain. A message like that appears 'No LIF on node "" ...' with no node name between "" because, obviously, there's not yet a second node.
Is there a way to bypass this?
I know that I could create an NFS datastore where temporarely vmotion the content but LAN is very slow and customer prefer FCP.
It's probably better if you set up the system as a single node cluster so you should be able to use FCP without many issues. I'm not too sure what you mean with "ghost node": did you join the second node and then just powered it off? If so do you remember if you enabled Cluster HA before shutting it down? I'm thinking that the second node missing is what is causing the problem here.
the issue has been solved. But first I would clarify something.
We have a 2554HA with both nodes in 7-Mode, let me call A and B. We've moved on the A aggregates the data in the B ones to free the disk shelves for a fresh setup of cDOT on B.
We've disabled the HA at the A node that now is running in 7 mode and serving users and apps (mostly VMware datastores)
In this way we've been able to setup the first node of a (future) 2 node cluster on B that has gone up and running fine waiting for the A (the ghost node) until the data will be migrated to its aggregate (we need to resume the production on B before to perform a fresh cDOT on A (and then join to the cluster).
Everything went good untile we've created the FC SVM. It was impossible because in some way (a wrong one) "knowing" that it was on a 2 node the SVM pretended the FC LIF on the second node too. But this second node was not yet present (the gost...)
It was then easier than expected after a few thoughts about. I've disabled everything linked to a 2 node clusters, so SFO and HA were disabled wiht a simple:
storage failover modify -mode non_ha -node “B”
In this way the cluster has been transformed in a single node one and the FC SVM could be build.
Once migrated the data from current A to B again we'll setup a fresh on A and then putting again "true" to the HA parameter, we will be able to join A to the cluster.
After all we've patented a way to transform a 2500HA FAS from 7 to cDOT one controller at time (having space of course to reverse data).