cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
696
Views
0
Helpful
1
Replies

Convert NX5K CE/vPC to Fabricpath/vPC+ with minimum distruption

Hi guys 

I am about to convert a pair of 5596UP+6x connected 2Ks from Classical Ethernet and vPC mode to FabricPath in production environment where hunders of servers connected and getting outage window is almost impossible! So you see my challenge its all about minimizing the impact.

I already red this:http://www.cisco.com/c/en/us/products/collateral/switches/nexus-5000-series-switches/white_paper_c11-709336.html and some other documents and completely aware of what actions need to be taken for peer-link port-channel , license , vlan mode etc.

I am going to it in this way : 

  1. Shutting down all server ports on 5Ks and 2Ks connected to secondary unit 
  2. Take down secondary unit by shutting down all port-channels (all uplinks north/south except Fex port-channels)
  3. 7Ks are already in FP mode so from them its just converting switchport trunk to fabricpath and removing vPC for links going to secondary 5K
  4. Convert all VLANs and uplinks to 7Ks to Fabricpath - Secondary Unit and configuring FP ID and E-ID etc.
  5. Convert Peer-link (which is down currently) to Fabricpath 
  6. No shut all over place on the secondary unit and Shut same interfaces on Primary unit (same as step 1 and 2)
  7. Run Step 1-5 on primary unit while it is down and join it to fabricpath peerlink with new setup

I know this is not like Cisco recommended way and not mentioned anywhere on any document but like I said taking down both chassis is not an option for me.

As you see on the step 6 there is a moment where I have one pair talking FP on the peer-link and uplinks to 7K and the other pair still in regular vPC until I shut the primary unit. 

Finally my question is , first of all how 5Ks will react to this kind of setup , more specifically what is the impact ? 

and secondly does it make sense to do it in this way ? any idea any recommendation is highly appreciated 

- Info: Servers are all connected to both units whether directly or to 2Ks 

-Info: 2Ks are one side only so each Fex is only connected to one pair either Primary or Secondary 

 

Thanks in advance 

Ehsan "Shawn"

 

1 Reply 1

## Notice : Do NOT try the above the procedure even in your LAB ! 

 

So turns out it might cause a hardware failure !! yes one of my 5596UP devices is dead no CLI no output , tried multiple cold reboot or PSU changeover !! its completely dead switch going to RMA the device !

Lucky me I was trying it in the LAB env. not in the production , anyway this is HOW to Brick a 5596 running nx-os 5.1.3 :

 

At the step 6 from above post , I did the no shut on the recently converted 5K to fabricpath , right away lot of memory errors flooded the screen related to forwarding engine and other process fwm etc. 

before even I could put any command it went to reboot and stopped right there ! blank screen on Console port !! first I thought its my serial cable then I noticed the FAN indicators lights are offline though fans are still spinning !! cold reboot , change PSU or FANs didn't help so it was a damage right to the hardware !! 

 

Again do not try it at any environment !! TAC has no explanation and no way we could bring back the device live so RMA in place ! 

 

YAY ! I broke a 5596UP in hardware level with just few commands in CLI !! Credit to me 

 

Cheers

Review Cisco Networking for a $25 gift card