Rename ACI Fabric Name?

[KVS7]

Anyone know how to rename the Fabric name?

During setup, you're required to enter a friendly logical fabric name. I incorrectly named it and can't rebuild it because I already registered a leaf/spine and that rebuild process would take much longer than hopefully just renaming the Fabric.

[RedNectar]

Hi @KVS7 ,

The process to rename a fabric is fairly simple to describe, but painful to implement.

To rename a fabric you simply wipe your system and start again!

If all you have done is registered some leaves and spines, the process should not be too painful. If you have gone further, then it is a little more painful!

[KVS7]

Will the APICs still see the leaf and spines if I rename the fabric? When I run "acidiag run lldptool in eth2-1" it says "Cisco Fabric Name TLV (different than the APIC) so I'm guessing the LEAF/SPINES might not be visible once I rebuild it with a different Fabric name.

[RedNectar]

Hi @KVS7 ,

If you have completely wiped the system (as explained in the link above) then once you've been through the discovery process the next time, the leaves and spines will learn the new name and will be visible again.

If you try and do this without following those steps carefully, then you WILL have problems.

[TiP: Before you wipe the system, record the serial numbers of the leaves/spines so you can allocate the same IDs next time. You can find the serial numbers via Fabric > Fabric Membership or use the show switch  command at the CLI]

BTW - the process for changing the VTEP IP address range is the same - even if you are just changing the mask!

[KVS7]

So I went ahead and rebuilt the primary APIC and it does not see the leaf/spines. So the fabric name is one of the parameters it uses to register a pending leaf/spine. What do I have to do now, wipe the leaf and spine or can I just wipe again, register the leaf/spine in order to wipe it? I don't currently have local access to the leaf/spine (different country). Therefore, would I be able to go back, register them in order to connect and wipe them or must I do it via local console cable? 

[RedNectar]

Hi @KVS7 ,

Looks like we were both at the keyboard at the same time - but yes - you DO need to wipe the laves and spine (as the link above explains). Now, if they are in a different country, you CAN do the process via SSH sessions to the OOB management IP address of the leaves and spines - assuming you got that far in the initial deployment. If not, then terminal server plus console cable is your best remote access option for the leaves/spines, and to access the APICs, you can use CIMC access - although you seen to have that covered.

[KVS7]

Thanks man, but real quick, I forgot to try and manually add the leaf/spine manually. Should I go back and try that first before wiping the leaf/spine? Actually I will try that real quick. I have the serials

[KVS7]

I was able to add it, but it says undiscovered with no register button. Would I be able to force a registration via command shell?

[KVS7]

Is my only other option to (fly to another country) console to the leaf and spine and run this procedure?

1. Physically disconnect all undiscovered leaf or spine switches from the APIC and any other leaf or spine switches
2. Login to the CLI of the undiscovered leaf switch and clean the switch by running the command "setup-clean-config.sh"
3. After the command completes, run vsh -c "reload"
4. Repeat steps #2 and #3 for other undiscovered leaf or spine switch
5. After all the undiscovered switches have been cleaned and reloaded, physically connect back the leaf/spine switches to the fabric

[RedNectar]

Hi @KVS7 ,

You won't need to physically disconnect anything, especially if you have the serial numbers to be able to re-assign the correct IDs.

The idea is:

1. When you run setup-clean-config.sh on each switch, you've put it into a "ready to reboot" state - but when it does reboot it is going to try and retrieve its config - and if the APIC is up and running when you reboot - well, bad things can happen
2. Once all the switches have their config cleared, clear the APIC - acidiag touch clean - TBH I don't think running acidiag touch setup does anything extra - but now the APIC is in a "ready to reboot" state - of course it too will re-learn the config if rebooted before all APICs are in this state.
3. Once everything is in a "ready to reboot" state - reboot them. Power cycle or the reload command (or acidiag reboot on the APICs)

Once the APICs are up, do the registration process with the correct fabric name, then the switch discovery should begin...

[KVS7]

Ok thanks, just to mention, the APICs have already been blown away with the correct fabric name. I guess all I need to do is clear the one leaf and spine.

[RedNectar · Accepted solution]

Hi @KVS7 ,

One leaf and spine? So this is a tiny lab setup I guess. If so, seems hardly worth the trouble of flying to another country to change the fabric name.

But, to be sure, I'd clear the APIC (APICs?) again after clearing the switches and before rebooting the switches.

[KVS7]

Not a lab, I setup a datacenter out of the box. The APIC only sees the first leaf switch, if you attempt to register it, it will then see just one spine but I rebuilt the APIC first before that spine can enumerate the rest of the leaf switches. Then the new appliance vector AV parameters didn't match the spine (like the UUID, fabric name I guess and the TEP address) and therefore I could no longer see that leaf and spine. So my guess is that I need to blow away that leaf and spine, but they are physically connected so I guess I would have to perform this procedure on both at the same time from the sounds of it. But only after I rebuild the APIC using the old parameters so that the leaf and spine are visible again and allow me to connect to them. 

[KVS7]

I rebuilt the APIC with the original settings so I can see the leaf switches now, but I get the same error I got before regarding code F1410 so I will end up rebuilding at location or making another post. This question has been answered. Running acidiag run lldptool in eth2-1 and eth2-2 will show what the leafs have received from a former APIC build. Since I removed the node from the controller, eth2-1 (active NIC) just went down hard and isn't coming up. According to the decommission/recommission guide, I need to run the setup-clean-config.sh <firmware image> command to default the port settings (just guessing this is why it went hard down)

[RedNectar]

Hi @KVS7 ,

OK - I get the reference to "one leaf and spine" now - that's as far as you got with the wrong name,

Looks like you've had to spend a lot of time in the docs. That can be pretty daunting, but hopefully a useful learning experience.