[CDOT] Change management IP addresses to different network

[BOIWABOIWA]

Hello,

I need to change the management IP addresses of our CDOT system; the new addresses are in a different network. Since we're currently using port e0a for management, I also want to switch back to e0M. Since there are quite some steps to achieve all this, I made a list of steps I have to perform:

1. Change IP-Address of SP on each node: system node service-processor network modify ...
2. Create routing groups for the new network for cluster and both node Vservers: network routing-groups create ...
3. Add e0M to the failover groups for cluster and node management: network interface failover-groups create ...
4. Change the network settings for cluster and node management: network interface modify ...
5. Remove e0a from the failover groups for cluster and node management: network interface failober-groups delete ...

 Did I miss anything?

Regards

[jt52]

I installed Cdot 8.3.2 on 3240 - during cluster setup trying to use e0M for cluster and node management. When I do I cannot access. If I use e0a I can access just fine, but if I change either the cluster or node management back to use e0M I no longer can access the cluster or node from ssh or https. Is there something else that has to be done or configured after the cluster setup is complete?

[xcaliber]

If you're setting up a new cluster then you should be specifying the e0M port during initial setup in the CLI.  I would run the cluster setup again and select the management interface to use e0M. The 3240 may use e0P or another interface potentially?  You can use whatever interfaces you want for management, but e0M is recommended as the primary. 

[SeanHatfield]

Make sure that the upstream switchport for e0M is configured as an access port in the correct VLAN, and that the switchport supports and is configured for 10/100 speeds.  On the older platforms the e0M port was not capable of gigabit, and many datacenter switches do not support 10/100.

[jt52]

Hi,

Ports on switch are set for access ports and 10/100, NetApp shows auto/100. When I unplugged the cable from management port both cluster and node management failed over to e0a, so I can access OnCommand. I reverted back to e0M and no longer connected. I dont have any vlans configured on NetApp yet.

Thank you

[jt52]

Hi,

I run cluster setup, it ask for cluster managemnet port, I type e0M, the IP, subnet mask, etc. It then ask what port to use for node management I again type e0M and enter another IP, same subnet. After setup finishes if I run network interface show it shows vserver name, lif name for cluster mgt and node mgt shows ip's I used and e0M port. all ports are up. But when I try to connect using either IP with ssh or https it does not connect. If I run the setup again and instead or e0M I use e0a I can connect just fine. Is there something on the switch that is configured different for these ports? anything I need to check in NetApp?

Thank you

[xcaliber]

Try taking a laptop and directly connecting to the e0M port on the same subnet (Assign static IP to laptop).  Check speed settings and see if you can ping the e0M interface.  It sounds like a switch configuration issue.  Are you able to determine which "Home Port" the e0M is assigned to?  2-controller cluster it should show you which controller currently "owns" or "homes" that port.

[jt52]

Hi,

I connected laptop, to managemnt on Head (ctrl) A and I can ping the network mgmt port e0M  IP on Head (ctrl) B. When I removed the cable for e0M it failed over to e0a. The home port is on controller A e0M but failed over to same controller e0a. When it failed over to e0a I cna log into OnCommand. I reverted back to home port and no onger have access.

Thank you

[xcaliber]

Ok that is a bit odd.  It should be failing over to the same port on the Controller B (e0M).  If you were able to ping it from the laptop - were you also able to access the management interface?  If it was responding to ping you should be able to load the MGMT interface on it.  Do you have cables connected to both e0M interfaces on both controllers?  Is the e0a interface the primary home port on controller B?  Via console command line can you re-home the port to different interfaces?

[xcaliber]

For cDOT 8.3.  I had some issues finding the correct command with NetApp documentation (annoying).  You can change the e0M network interface IP/subnet with the following command (Run this from console connection)

CLUSTER::> network interface show -curr-port e0M

displays entries that can be modified for the e0M management port for proper command syntax

CLUSTER::> set advance

advance command structure - approve

CLUSTER::> network interface modify -vserver VSERVER -lif LIFNAME -home-node HOMENODE -home-port e0M -address IPADDRESS -netmask NETMASK

you may have to include -force-subnet-assocation

The subnet should be created prior to moving IPs from my experience.

[shyfur]

Hi,

Could you please tell me how I can change node mgmt IP and cluster mgmt IP to a different IP subnet? Please tell me the procedure.

[nigelg1965]

I'd like to know too.

Seems this question baffles even NetApp's own staff for the response I've got so far.

[SeanHatfield]

"It Depends".

No, really.  Back in the days of 7mode you just hand edited a few config files and rebooted.  But cDOT is database driven, and the best way to get there depends on your situation, how many lifs need to be re-addressed, can you connect to the old subnet and the new subnet at the same time, and are you on a release from before or after the introduction of broadcast domains.

At a high level I'd suggest configuring ports with connectivity to the new network, creating new lifs, etc on those ports for node mgmt and cluster mgmt, and then decommisioning the old lifs/ports.  It's likely non-disruptive if planned correctly, but the steps to get there will vary from one situation to the next.

[nigelg1965]

Thanks

This is a 2254, so connectivity is limited, the LIFs etc are the ones the installation wizard creates.

I think I can find away to do it by temporarily designating an unused 10GB port on each head as a node management one, deleting the e0M lifs and create new ones in the right subnet.

Really this post is about NetApp inability to offer a solution to what can't be that unusual a question. Last time I had a similar one they ended up recommending to wipe the filer and start again, Which I did as there wasn't really any other set-up it.

Crux of the matter is CDot may be a great product, but it may be so complex now that NetApp's staff can't really support properly.

[mbeattie · Accepted solution]

Hi,

Also consider:

6. Updating DNS with the appropriate A & PTR records (And CName aliases if you have them configured)

7. Updating any firewall rules in your environment

/matt