AP Maintenance Cycle

[Ronald_Dvorak]

https://extremeportal.force.com/ExtrArticleDetail?an=000089011

No sure why someone want to reboot his APs every day but here my scenario.

Upgrade a controller pair to a newer release - set APs to controlled upgrade - so the APs don't upgrade to the new version and there is no service outage.
Then disable controlled upgrade and enable "AP Maintenance Cycle" so the APs reboot during the night (off-peak hours) to load the new software.

I've tried it in my lab and I don't like/unterstand the behavior.

In my lab setup I have only 15APs and only one with clients connected.
So in "phase#1" (= reboot APs with no connected clients) 14 of the APs reboot.
Before the APs were loaded with the new image and up&running also the AP with clients connected started the reboot.

Could you please explain it in more detail.
I.e. why the duration of 1-6 hours.
A AP need about 2 minutes to reboot and connect again to the controller.

Let's assume a C5210 with 900APs.
In phase#1 (no clients connected) let's say 1/3 have no clients and restart.

Then without any waiting time phase#2 is started and 100APs are in this category and reboot.

Next phase#3, my channel plan is 1/6/11 which would mean 3 groups with a 2minute waiting time between the groups.

Result is that all APs have restarted in timeframe of 10minutes.

My expection was that if I set the duration to 1hour that for example 15minutes are reserved for phase#1 - if some clients roam to another AP and so the AP has no clients connected.
Then another 15minutes to wait whether a AP would fit in phase#2 and the rest is for phase#3.

The conclusion is that this function is not useful in my scenario.

Or is my setup to small to see the "real" behavior of the feature.
Any whitepapers that explain the details/timers.

Thx,
Ron

[Ronald_Dvorak]

James, I've just done the test, you are right controlled upgrade loads the firmware and in the meantime provides service to the client.
So the downtime during the reboot + DFS was 2:20minutes.

Would be cool to combine controlled upgrade + AP Maint Cycle so I'd avoid this downtime for all APs at the same time.

The method that I use right now to avoid it is to controll upgrade only every 2nd AP which is a lot of double-tap tab and hit space and do that 400 times for a customer with 800APs.
Or even smaller groups if necessary.

I always thought only Cisco APs need a reboot every night to run stable 🙂

-Ron

[rickrey]

Does anyone know if this combination of AP Maintenance Cycle and controlled upgrade works as suggested by Ronald?
I got a better feeling if I know, just the chosen APs get updated during the maintenance cycle.

[Ronald_Dvorak]

Thx Doug.

For my scenario the sleep timer is too low as it would take more time till the AP could FTP the image, upgrade, reconnect, bring the radio up and provide service.

As I've mentioned it was only a test with 15APs so I wonder how long it would take for a fully loaded C5210.
If I look on this data I'd imagine not very long so I don't get the duration of max. 6hours.
My assumption was that if I set it to 6hours that there is more waiting in between the steps to get a more smooth transition.

Could be that I don't see the value of the funtions as I never was looking for a feature that could reboot my APs every night ..... any real scenarios for that as I can't think of one.

-Ron

[Doug]

More details on the feature....

• Maintain all APs without any clients and remove these APs from list of APs to maintain, APs that are removed will restart
• Sleep 120 seconds
• Retrieve AP stats
• Compare the new wired traffic stats with the stored one and divide it by 60 seconds
• Maintain APs with traffic less than 1kb/sec and remove these APs from list of APs to maintain, APs that are removed will restart
• Sleep 120 seconds
• For the remaining APs, walk through their radio stats and group the APs by a combination of frequencies. With correct deployments of APs, each point of the coverage area will be covered by more than one AP on a different channel, so grouping APs by frequencies will ensure that during the maintenance window, the whole area should still be covered even when some APs are restarted.
• Maintain each group of APs and sleep 120 seconds after each maintain.

What other options would you like to see to minimize downtime of the WLAN-Service versus a force reboot of all the access points at once?