Hello, I would like to discuss another method of a bulk controller upgrade and see what other engineers take on this upgrade path would be.
Say I have an instance of 8 4404s with 50 APs each, In this case I have N+1 redundancy where I can follow the normal proceedure
- Move all APs to controllers 1-4
- Preload all APs with the new code version
- Upgrade and reboot empty controllers 5-8 to new code version
- Move all APs to 5-8 with new code versoin
- Upgrade empty 1-4
- Move all APs back home
Now take the same scenario only chage it to 80 APs per controller. I've now lost my N+1 and cannot do it quite as smoothly.
As opposed to trying to follow the normal proceedure and have an extended window of "brown outs" How about doing it all at once.
Black-out accelerated proceedure:
- Preload new code on all controllers
- Preload new image on all APs on all controllers
- Reboot all 8 controllers at the same time.
- Allow time for APs to connect back and load the new image
I assume with this proceedure that I might see around 15-30 minutes of actual downtime to the site but it seems like that could be preferable to two-three hours of brown outs.
What are your thoughts, and do you feel that 15-30 minutes is
Welcome to CSC!
I have 40+ WLCs and like you looked at this very closely. In my environment I do a predowload and reboot them all. Ive tried the surgical approach and in my case if you get aps bouncing back and forth on WLCs they will upgrade then downgrade etc.
Also note -- If you are on 188.8.131.52 code, predownload "may" have a bug. I noticed an issue during a recent upgrade in mu network and later someone commented about the same issue here on CSC.