Change in PoE consumption on Access Points after updating WLC

[f0rslund]

Hello!

After our latest update of our WLC we have noticed that AP's now reserve full power from PoE switches.

The biggest problem we have is that we have some SG-300/350 switches with 6-8 2700/2800 AP's connected which in reality draws max 12 watt each now have problems because it looks like the AP now reserves 22.5 watt and maxes the PoE reserve of the switch.

The problem it causes is that 1-3 AP's which used to work on the same switch now won't get any power.

I have tried to manually limit the PoE output from the switch on each interface, but because we are in this scenario working with the fantastic SG switches that doesn't seem to exist..

Is there an easy solution to this?

Thanks in advance!

[Leo Laohoo]

On the 2800, disable the USB port.  It will save a few watts (per AP).

[balaji.bandi]

check what is the configuration on the port - is this auto, then adjust reservation as mentioned below guidelines :

https://www.cisco.com/c/en/us/td/docs/routers/nfvis/switch_command/b-nfvis-switch-command-reference/power_over_ethernet_commands.html

[Arshad Safrulla]

Both of them are POE+(802.3at) rated devices, and require POE+ to get all the advertised features working. It can work in POE(802.3af) rated power, but there will be limitations in certain features. You may refer the datasheet of both AP's to know what will be the limitations of using different POE standards. 

And also you should keep in mind that the what you see inside the bracket is the budgeted power for that port, it doesn't mean that the device connected to that port going to consume the same amount. Real world consumption depends on antenna gain, number of clients connected etc.

That being said it is never recommended to over subscribe POE budget of a switch. If your switch is only rated for 120watts, I would say try to limit yourself to maximum of 5 AP's per switch. Just because an upgrade I do not think AP's magically increased it's POE consumption. As @Leo Laohoo suggested try to kill the USB port, additionally check whether you can hardcode the power delivery to 15.4watts per port, this way you can get all the AP's working but all with reduced performance.

[David Ritter]

Interesting, my 2960XR's report:

2702/3702's 16.8 watts, 2802 26.1 watts, nd 3802 30watts  In some cases I had to swap out 3802's for 28's because I could not get that 30w at the end of the cable.

[Scott Fella]

You should always make sure you look at the power budget consumption using the power draw defined on the access point data sheet.  This will eliminate issue like what you are running into.  If after an upgrade, this happened and you lost ap's due to the ap now requesting the power defined in the data sheet, your best bet is to downgrade back, get you environment up and running and plan for your next steps.

[Rich R]

While what Scott says is true about checking data sheets Cisco managed to make a real mess with the 9100 APs by incorrectly stating  the POE consumption of the 9100 APs on the data sheets.  We got caught by that because we planned before we had any to test and had to revise when we got hands on APs and discovered the data sheets were wrong (quite significantly):

9115 - PoE+ 20.4w according to data sheet actually using 24.1w
9120 - PoE+ 25.5w according to data sheet actually using 30.0w

Cisco opened an obscure (because it highlights the data sheet only mentioning LLDP not CDP which is the Cisco primary method) documentation bug: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCvt91990

This was more than a year ago and looks like the data sheets still haven't been updated - anyone feel like submitting feedback to remind them to fix the POE info?

There have also been some bugs related to POE changes in different releases.

So the old maxim applies - always test and verify before deploying.

[Roger]

We ran into the same problem when we started migrating APs from 5520:s to 9800:s.

All of a sudden APs that hade been working flawlessly for years started reserving waaay more power than they ever use resulting in scores of APs never getting any power whatsoever from the switch....

The problem is most apparent for 2802:s etc.

The only "solution" we found to get the APs up-and-running was to disable LLDP and CDP on the AP-ports, which f'ing sucks....

ie.

interface Gigx

no lldp transmit
no lldp receive
no cdp enable

NOTE! After this change you have to do a "power inline never" -> "power inline auto" on the interface for it to release the reserved power, the log says "... until port down/up cycle" but a reboot of the AP or "shut/no shut" on the interace won't do it.

[Scott Fella]

In production, I only use CDP and not both.  We ran into issues with ap's not negotiating properly when CDP and LDDP was enabled on the switch port.  Once we disabled lldp, we stopped having ap's bounce.

[Leo Laohoo]

Only Catalyst switches that are running IOS-XE have CDP/LLDP issue.  This issue was never present with switches running classic IOS.  

There are multiple public-facing Bug IDs, such as:  CSCvt33980, CSCvv54912, CSCvr24032 (enhancement request), etc.

[garagerevolumark]

Both of them are POE+(802.3at) rated devices, and therefore both require POE+ to function properly. It will function with POE(802.3af) rated power, however some features will be limited. You may look up the limits of utilising various POE standards on the datasheets for both APs.

Also, take in mind that the budgeted power for that port shown inside the bracket does not necessarily imply that the item attached to that port will consume the same amount. The amount of energy consumed in the real world is determined by antenna gain, the number of clients connected, and other factors.

Having stated that, it is never a good idea to exceed a switch's POE budget. If your switch is only rated for 120 watts, I recommend keeping the number of APs per switch to a maximum of five. I don't believe the AP's POE usage has miraculously increased as a result of the change. Try to disable the USB port, and see if you can hardcode the power delivery to 15.4 watts per port, as proposed by @f0rslund. This way, you can get all of the APs to operate, but with lower performance.