ONTAP Discussions

Removing Nodes From a Cluster


We are planning on removing a couple of HA pairs from a cluster, then of cause correspong LIF's on these nodes as well. 

The problem is that these LIF's are hard coded on clients for mounting. For instance, they used SVM-lif1 for node1 and SVM-lif2 for node2  to reference volumes. When we remove these nodes, LIF's will be removed as well.


One solution I can think of is to create DNS alias for these legacy LIF's. the problem with that is these legacy LIF's would be staying there "forever". 


Are there any better solution for that?



Are the LIFs for ISCSi or CIFS/NFS?

CIFS & NFS LIFs can be migrated to other nodes and it is *generally* non disruptive. This happens during failovers/givebacks during updates, so if your workload isn't interupted during ONTAP updates, it should be fine.


ISCSi LIFs should be multipathed? But then you wouldn't be using a single ip address.....

First of all. Hard coding is bad.

Second of all, NAS LIFs you can migrate online.
While it is possible to migrate SAN LIFs but only while it is offline, so in this case the compromise might be either rewrite your code or expect some small outage.

Understood that LIF’s will be failed over to the other nodes and there will be no interruption.

Since they are using svm-lif1 as the the nfs server in their mount configuration, what will happen next time when they are trying to remount? If changing the name of svm-lif they are using on clients is the only option here? I am trying to avoid it, because there are a lot of such clients.


An NFS SVM lif is not hard linked to a node, you can move it to whichever node is appropriate.

If you're getting rid of the SVM, then that's a bigger problem.


Also, migrate the LIF manually before hand, and modify it so that it's home port is on the new node.

Understood all you are saying.

I am concerned about client side. They are using svm-lif1 now. After the node1 is removed this svm-lif1 will be no longer there. What will happen next time when these cluents try to remount by using the same svm -lif 1? To solve the issue, Do we have to change it to the other existing one on all these clients, svm -lif6, for instance?

Hope I made myself clear this time


The SVM Lifs (for NFS) are not tied to a node.


You do a network interface migrate  and move it to another node.


It will require that the port on the new node has the same network VLANs etc, but the client should never notice.


for example:

network interface migrate -vserver <vservername> -lif svm-lif1 -destination-node <new-node> -destination-port <destination port>



Regarding this sentence: " After the node1 is removed this svm-lif1 will be no longer there."

This is not definitively true. You can move svm-lif1 to a different node. This is what OZWALKERZ is saying. If you follow his instructions, you can move the LIF to another node and the clients can keep connecting to it as they are now.


I got it now. My appology! I didn't get his points. That means a node can have multiple LIF's for a SVM. 

Thank you all, specifically OZWALKERZ .