No MGMT LIF for 2nd node of cluster

[wet hinge]

Hi all, hope everyone is having a good week.

Yesterday I was utilizing the cluster setup wizard to complete the initial setup for a new 2 node AFF-A150. I chose to complete this on the CLI rather than the System Manager GUI. I began on Node 1 by giving the controller it's mgmt LIF information and creating the cluster. At the end of this, I was given "to complete cluster setup, you must join each additional node to the cluster by running "system node show-discovered" and "cluster add-node" from a node in the cluster". According to ONTAP documentation I was under the impression that at this point in the initial setup I would hop over to Node 2 via console cable, open up a new session and go through the cluster setup wizard on that node. I know there is a point where it asks if you want to create or join another cluster, thats when I have always just pointed it to one of the 1st nodes cluster IP's and joined the cluster that way.

This time I decided to follow the instructions that the CLI was telling me, and ran "system node show-discovered" and "cluster add-node" from Node 1. It found Node 2's cluster IPs and I added it without issue. I ran "cluster show" and the 2nd node showed up healthy and eligible so I figured I was now done with the cluster setup wizard portion and I could move on to actually configuring the system. However upon running net int show I found no mgmt LIF for my 2nd node. I looked as best I could, also checking the GUI to see what it said and I indeed did not have a mgmt LIF automatically created for my 2nd node, as had been done for my 1st. It was easy enough to create a mgmt LIF for Node 2, I made sure it was in the same broadcast-domain as the other node mgmt LIF and gave it the proper home node/port. I was able to successfully SSH to it as well. There isn't really a problem now - I'm just wondering what happened? Is there any cause for concern since it wasn't auto-created?

[strange belfry]

The "normal" way to create and join a node is from the nodes from the CLI themselfs.. so from node 1: "cluster setup" where you setup node mgmt, and then cluster IP... from node 2: "cluster setup" where you setup node mgmt and are asked for an IP of node 1's cluster IP (not mgmt IP)... the node is then added to the cluster... I have never dont it like you did (join node 2 from node 1).. 🙂 I normally use a serial connection or the service-processor to do all this.. You should be able to setup the node mgmt IP manually woth something like "net inter create... pointing to the e0M port..."

[wet hinge]

Right, you describe the way that I was intending to do it, however either the wording in the cluster setup wizard is difficult to properly understand or it should be telling the administrator to do it the normal way. I created the node mgmt LIF as you mentioned, so unless anyone knows of the way I did it causing problems I guess I will just move forward with the config. Thanks!

[strange belfry]

You might be right. I think I have been doing this for way too long to even read the words 😉

[lilac swallow]

Dumb question, but do you have the cluster setup wizard wording? I don't set up enough clusters being a perf resource in support to be familiar with it. I wonder if we have room to improve it?

[wet hinge]

It was "to complete cluster setup, you must join each additional node to the cluster by running "system node show-discovered" and "cluster add-node" from a node in the cluster"

[lilac swallow]

Are you in a position to open a case?

(not today, just at first available time)

[wet hinge]

Yes, I could open one tomorrow. Is there some concern? The LIF is there now and seemingly working normally

[lilac swallow]

No. Literally just say "Paul Stejskal suggested in Discord to open a case to file a bug. The bug is to change the wording to do the cluster join instead of what it says now..."

[wet hinge]

Roger, can do! I will link the ONTAP 9 documentation where it describes what @strange belfry suggested.

[lilac swallow]

Great.

Actually call it RFE not bug.

It's technically RFE.

(request for enhancement)

[wet hinge]

https://tenor.com/view/imdb-comic-con-imdb-comic-con-gifs-kevin-smith-learned-something-imdboat-gif-9790411

[turbid cliff]

So, from my experience
(I only use the CLI, and never the GUI)
I setup the first node of the cluster (cluster setup and follow through to end). Then I go to the second node and do the same. I need to provide a Node-MGMT address and one of the CLUSTER-LIF IP addresses. Usually finishes just fine, and the cluster is in HA mode when complete (2-node)

When I try the other way:
Setup the first node to completion. I still go to the second node and run the basic "cluster setup" to t a minimum create the node-MGMT and establish the CLuster LIF IPs. THen from the first node, use the "cluster add-node" and specify one of the CLUSTER IP addresses.
I have notied using this method DOES-NOT turn on Cluster-HA. After the node is ready, I need to do "cluster modify -ha true"

[wet hinge]

In the method I used above it also breaks HA, I had to do the same when I was done.

[turbid cliff]

It doesn’t really break ha in as much as it is not automatically enabled

[sharp locust]

BTW: The official docs says to join the other nodes FROM the other nodes (just like everybody does): https://docs.netapp.com/us-en/ontap/software_setup/task_join_remaining_nodes_to_the_cluster.html
No "please-use-cluster-add" from the first node where you created the cluster. 🙃

[wet hinge]

Yeah that’s why I made my post! I got curious when I saw the wording in the CLI and decided to try it that weird way - will definitely be staying with the standard way of Node 2 joining cluster from Node 2 moving forward.