how to switch ifgrp ports from 1Gb to 10Gb in 9.1 ?

[atwilson]

Hi

Is there a recommended procedure to switch the ifgrp ports from 1Gb to 10Gb in ONTAP 9.1 ?

In the olden days of 7-mode it was a relatively simple procedure edit rc files and failover the nodes etc.

Not sure how to approach this in ONTAP 9.1.

The cluster nodes currently have multimode lacp ifgrp's with 2x1Gb ports each.    We are introducing 10Gb infrastructure at that site and want to switch the ifgrp's over to the 10Gb ports.

Thanks

[SpindleNinja]

Are you reusing the same ports or using different ports? 

But moving it over is even easier than 7mode. 

[atwilson]

the 10Gb ports are different ports

[SpindleNinja]

the migration would go something like this: 

1. Connect ports and create the new ifgrps and tag vlans if needed.  (each of these objects become ports for you to put lifs on) 
2. Move those new "ports" to the correct broadcast domain(s).  
3. Move the lifs to the new ports 
4. Remove 1 G ports from broadcast domain(s) 

[atwilson]

aargh!  was hoping I could do something clever like migrate the N1 lifs to N2, add the 10Gb ports to the existing N1 ifgrp and remove the 1Gb ports, then migrate the lifs back to N1.  repeat for N2.

[SpindleNinja]

I believe that should also work,  I've found just adding, moving and removing easy.   Just be sure to not have lifs on whatever port you're working on. 

[atwilson]

<relief> glad to hear you say it should work.  how confident are you?  I have not been able to find a related NetApp article or discussion to confirm if it will.    I would prefer to avoid creating new ifgrps and vlans and all that reconfig.  we have many vlans and lifs.

[SpindleNinja]

It's really simple and straight forward,  I actually find 7mode networking to be cumbersome when trying to add and remove things.  

Here's a doc that states "do not mix ports in ifgrps"  https://www.netapp.com/us/media/tr-4182.pdf   

But I did test this out in my lab and was able to add both a 1G and 10G port in to the same ifgrp.    However, I don't have switches that do both 1 and 10 G ports so I can't actually built a live ifgrp.   

CLUSTER::> net port ifgrp show -instance
(network port ifgrp show)

Node: CLUSTER-01
Interface Group Name: a0a
Distribution Function: ip
Create Policy: multimode_lacp
MAC Address: 02:xx:xx:7f:xx:xx
Port Participation: none
Network Ports: e0b, e0d
Up Ports: -
Down Ports: e0b, e0d

a0a Default - down 1500 auto/- -
e0b Default - down 1500 auto/10 -
e0d Default - up 1500 auto/10000 healthy

like I said,  I think it might work,  I just wouldn't do it in production. 

[atwilson]

thanks again, and yes I have reviewed that document along with the NetApp documentation for the network port ifgrp commands.    Fortunately  this particular change will occur during a scheduled maintenance period because the rest of the infrastructure (networking and hosts) will be down for their corresponding 10Gb upgrades too.

My dilemma is to determine if it is acceptable to move the lifs off a node, swap the ports in the ifgrp via the add/remove commands, confirm the ifgrp is up (with 10Gb ports only), move a lif or two back to check if it is working before moving the rest of the lifs.    Afterwards I will probably do a failover/giveback of each node to see how they boot

[SpindleNinja]

One of the advantages of Clustered ONTAP is the ability to evacuate  lifs off a port for migration or maintenance.  Thats why there is a 20G or more of cluster networking connecting each node.  

You can have an ifgrp without active ports,  so yes you could remove the 1G ports and add the 10G ports,  but ONTAP will see the port as down and migrate the lifs to other port(s) in the failover group.   I don't like to keep lifs on ports i'm messing with that's all, but if the switches are all going down,  it don't matter. 

net int migrate-all -node <node> -port <port>  will migrate all interfaces off the port to the failover port(s) manually.   

and "net int revert *" to send everything home when you're done.