Switching

We have a stack of four EX4200-48PX switches, running Junos 15.1R7.9. One of these switches has 31 Ruckus R710 access points patched in, the remaining 17 ports have nothing patched in. Yesterday we had a power outage and the entire stack went down after the UPS ran out of juice. Two hours later the power came back on, stack started up and now I can only get 24 ports on this switch to deliver PoE. The PoE firmware is upgraded to the latest. Below is the output of "show poe controller", the switch in question is number 1 which is only drawing 134.20 watts.

 

Controller index	Maximum power	Power consumption	Guard band	Management	Status	Lldp Priority
0	740.00W	69.00W	0W	Class	AT_MODE	Disable
1	740.00W	134.20W	0W	Class	AT_MODE	Disable
2	740.00W	41.20W	0W	Class	AT_MODE	Disable
3	740.00W	126.80W	0W	Class	AT_MODE	Disable

 

Below is the output of "show poe interface". You will notice that ports 0 through 24 are putting out PoE power and ports 25 to 31 are not. Any ideas?

 

Interface	Admin status	Oper status	Max power	Priority	Power consumption	Class
ge-1/0/0	Enabled	ON	25.0W(L)	Low	5.2W	4
ge-1/0/1	Enabled	ON	25.0W(L)	Low	5.2W	4
ge-1/0/2	Enabled	ON	25.0W(L)	Low	5.0W	4
ge-1/0/3	Enabled	ON	25.0W(L)	Low	5.6W	4
ge-1/0/4	Enabled	ON	25.0W(L)	Low	5.0W	4
ge-1/0/5	Enabled	ON	25.0W(L)	Low	4.8W	4
ge-1/0/6	Enabled	ON	25.0W(L)	Low	5.0W	4
ge-1/0/7	Enabled	ON	25.0W(L)	Low	4.8W	4
ge-1/0/8	Enabled	ON	25.0W(L)	Low	4.8W	4
ge-1/0/9	Enabled	ON	25.0W(L)	Low	4.8W	4
ge-1/0/10	Enabled	ON	25.0W(L)	Low	4.6W	4
ge-1/0/11	Enabled	ON	25.0W(L)	Low	4.6W	4
ge-1/0/12	Enabled	ON	25.0W(L)	Low	4.2W	4
ge-1/0/13	Enabled	ON	25.0W(L)	Low	4.2W	4
ge-1/0/14	Enabled	ON	25.0W(L)	Low	5.4W	4
ge-1/0/15	Enabled	ON	25.0W(L)	Low	4.8W	4
ge-1/0/16	Enabled	ON	25.0W(L)	Low	5.0W	4
ge-1/0/17	Enabled	ON	25.0W(L)	Low	5.2W	4
ge-1/0/18	Enabled	ON	25.0W(L)	Low	4.6W	4
ge-1/0/19	Enabled	ON	25.0W(L)	Low	4.8W	4
ge-1/0/20	Enabled	ON	25.0W(L)	Low	5.8W	4
ge-1/0/21	Enabled	ON	25.0W(L)	Low	4.8W	4
ge-1/0/22	Enabled	ON	25.0W(L)	Low	9.6W	4
ge-1/0/23	Enabled	ON	25.0W(L)	Low	4.2W	4
ge-1/0/24	Enabled	ON	25.0W(L)	Low	5.4W	4
ge-1/0/25	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/26	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/27	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/28	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/29	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/30	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/31	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/32	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/33	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/34	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/35	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/36	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/37	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/38	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/39	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/40	Enabled	OFF	30.0W	Low	0.0W	not-applicable
ge-1/0/41	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/42	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/43	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/44	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/45	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/46	Enabled	OFF	15.4W	Low	0.0W	not-applicable
ge-1/0/47	Enabled	OFF	15.4W	Low	0.0W	not-applicable

 

 

Hi HendersonD,

 

From this output, POE seems to be enabled on all the ports.  Are there any trunk ports on this switch? If yes, please turn off POE on the trunks (if any) and check.  May sometimes need to bounce the ports after the change.

 

Let me know if that worked.

 

Hope this helps.

 

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

We have fiber connections to the top and bottom switch in this stack through the EX4200 uplink ports. PoE is not enabled on them. Every port on this switch that has an access point plugged in is setup as a trunk port since we need to carry several vlans to each AP. In other words, I cannot disable PoE on this ports since the AP needs PoE and multiple vlans

I do not think your situation has anything to do with port configuration of Trunk or Access.  I did a lot of work with EX4200 and POE previously, BUT I only used 12.3 based code.

 

It looks to me like the switch/POE Controller "thinks" every port needs a reserve of 30W (class 4) despite LLDP-MED telling in 25W.  The reason I say this is 24 x 30 = 720, and POE Budget for EX4200-48PX is 740W max.  So if port 25 asks for 25 or 30W, no power would be applied.

 

Yes, this should NOT be happening -:(

 

Now since this all worked once, potential re-boot(s) of whole VC might clear it, but I might suggest instead you try with 12.3R12-S[latest] and see if results change.  No need to worry about 12.3, as this code will be supported on legacy (non-ELS) EX switches for same amount of time as 15.1.

 

Good luck.

We have a lot of stacks of EX4200 around campus and they are all running the same code at the moment. I am a bit reluctant to run a different version of code on this one stack. Is there anything else I can do to the switch to get PoE working properly? A little history might help here. Months ago we were running a previous version of 15.1 code. We upgraded to Junos 15.1R7.9 and never had a problem until this power outage. I did reboot the entire stack this morning and that did not bring back PoE on the ports in question. Why would PoE be working fine for months, we get a power outage that downs all of my switches, and when the come back up this one stack is having problems? I had a secretary call me a few minutes back and her IP phone and computer are not working. She is hooked to the same stack, different switch in the stack. Since her phone is not getting PoE I am thinking at least one other switch in this stack is having a similar issue

One other thing that might figure into the problem. Our IP phone system we got up and running just a few weeks back. Our switch stacks have never been rebooted since the new phones went in. We do have this on every switch

poe {
interface all;
}

 

I cannot figure out any relationship between the IP phones going in, the switches being rebooted and losing PoE power on this one switch that only has wireless access point plugged into it, but I wanted to mention this

 

 

Hi HendersonD,

 

The "set poe interface all" configuration should be it.  Re-iterating myself, what I meant is if you have any of the switch ports that are not actually connecting to APs/end-hosts that are POE devices, please disable POE on those ports.  Enabling POE on any other ports that simply connect to other switches for example, creates back pressure and POE issues for sure.

 

Hope this helps.

 

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

 

Can't help you with the why, but I believe (at least internally) switch thinks 30W per port, which allows only 24 ports active.

 

As for why, you're going to have to work with TAC, but I still think trying 12.3 might get you a solution faster.

 

Up to you, . . .

We do not have PoE enabled on any ports that connect to another switch. We have single mode fiber connections between switch stacks and our Juniper core in a hub and spoke configuration. There is one caveat to this, our Mitel phones (like all IP phones) have a two port switch on them. Ethernet patch cable from wall plate to phone. A second ethernet patch cable from phone to desktop computer or printer. So we are left with only a small number of ports that have just a printer or desktop (no phone daisy chained in between) connected that we could turn off PoE power.

 

Again, our phone system is brand new, I am guessing this one switch stack has 30-40 IP phones plugged in as well. None of these IP phones are plugged into the one switch that has all the access points plugged. This stack has been rebooted many times over the years as we upgrade Junos and never had a problem. I do not think it is a coincidence that the first time the stack is restarted after our IP phones go in and PoE is having a problem.

I personally do not believe in coincedence when it comes to Networking.  Have not seen it in 40 years I've been involved!!

Facts so far:

1. This switch stack has been in place for 6 years with multiple Junos updates requiring restarts. PoE has never been a problem
2. We have had for several PoE security cameras and PoE wireless access points plugged into these switches for years. Again, no problem even after a stack reboot
3. We installed a new IP phone system. The phones need PoE and have a 2 port gig switch built into the back of them
4. Many of these phones are daisy chained with printers or desktop computers. Patch cable from wall to phone, patch cable from phone to desktop or printer. Some phones have nothing daisy chained with them
5. We have PoE enabled on all ports on this switch stack
6. Juniper 4200-48PX should be able to provide plenty of PoE power to the items plugged in
7. We are running Junos 15.1R7.9 which is a JTac recommended version
8. The first time this switch stack is booted after the phone installation, we cannot get PoE on several ports that have been working fine for years

I have been doing this type of work for 20+ years myself and the phone/PoE connection is not a coincidence. Is there some other troubleshooting I can do or should I open a ticket through our local Mitel reseller/integrator?

Hi HendersonD,

 

A couple of things to check:

 

a) Are any of these interfaces physically up?

show interfaces terse ge-1/0/25

show interfaces terse ge-1/0/26

show interfaces terse ge-1/0/27

etc.

If there are any phones (especially ones with a switch) already up with external power, please try disabling POE on them and bounce the rest of the interfaces or toggle POE on them.

 

b) Please check the status messages from logs (check older file "messages.x.gz" if this has rolled over)

show log messages | grep poe | no-more

 

Note that some of these might be expected during the interface state change at reboot time, but helpful to check if anything special for the "OFF" ports.

 

Hope this helps.

 

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

Here is the output of a few different commands

root@HS102B> show interfaces terse ge-1/0/25
Interface        Admin    Link      Proto         Local       Remote
ge-1/0/25     up           down
ge-1/0/25.0  up           down   eth-switch

 

root@HS102B> show log messages | match ge-1/0/25
May 1 22:19:35 HS102B chassisd[1332]: PoE port name ge-1/0/25 status 60
May 1 22:42:55 HS102B chassisd[1332]: PoE port name ge-1/0/25 status 60

 

The command above is interesting. This is one of the ports that is not putting out PoE. Below is the same command from a port that is putting out PoE

root@HS102B> show log messages | match ge-1/0/24
May 1 22:19:35 HS102B chassisd[1332]: PoE port name ge-1/0/24 status 1
May 1 22:42:55 HS102B chassisd[1332]: PoE port name ge-1/0/24 status 1
May 2 07:29:50 HS102B chassisd[1332]: PoE port name ge-1/0/24 status 1
May 2 07:30:04 HS102B chassisd[1332]: PoE port name ge-1/0/24 status 1

 

Not sure what status 1 (this interface is putting out PoE) versus status 60 (this interface is not putting out PoE) means

 

We are thinking that our PoE problem is a hardware failure. We are opening a case with Juniper

Hi HendersonD,

 

Status 60 tells us that ‘Calculated power > power limit’.  Can you also check these two on the switch?

show log messages | grep poe_get_psu
show configuration poe | display set

 

Hope this helps.

 

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

 

 

@HendersonD, I know you don't really want to listen, but I think I hit the hammer on the head when I said the issue was calculated power.  I very much doubt this is a hardware issue, but something screwed up in software.  Now could be HW issue with POE Controller, but looks much more like software to me.

 

TAC needs to work this, but once again (and I know it might be painful) I suggest you load this VC with 12.3R12-S[latest] and see what you get for results.  I would think log file messages for any/all interfaces not providing power show "status 60", yes?

 

The above may be something TAC recommends, at minimum for troubleshooting data point, to start with.

Tonight I will load the 12.3 code and report back in the morning

I upgraded this switch stack to Junos 12.3R12-S12 which is the JTac recommended version. It did not make any difference, I still have the same ports without PoE power. I have another edge closet with only one EX4200-PX switch. This switch has nearly all of its ports requiring PoE between phones, security cameras, and wireless access points. It had two ports where PoE was working and now is not working. I also installed the Junos 12.3 code on this switch and it made no difference.

 

Several months back we had an EX4200 with PoE problems. We would boot the switch, all ports went active with blinking lights everywhere. We would watch as ports would start failing one after the other with lights going out.

When I first started troubleshooting this problem, I went to the command line and issued a show poe controller command. This is the first time I had ever used this command since we have never had PoE problems before. The message I received was the PoE firmware needed upgrading. I was not even aware that upgrading PoE firmware was a separate process from upgrading Junos. We applied the upgrade to PoE to all of our switches. I do not know what PoE firmware version that was installed, but it is now version 4.1.6. Could the PoE firmware upgrade be the root cause of this issue?

Hi HendersonD,

 

Absolutely, did you mean that message to upgrade the POE firmware was seen after the first reboot while on the Junos 15.1? That means the Junos version has a more latest POE firmware than what's installed on the switch at the time:
https://www.juniper.net/documentation/en_US/junos/topics/task/installation/poe-controller-software-upgrading.html

 

Assume the firmware upgrade has resolved it for you now? That's good news :).

 

Hope this helps.

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

 

We upgrade to 15.1 code a long time ago on these switches. At that time I had no idea there was also a PoE firmware upgrade that needed to be run. It was only last week when my switches rebooted due to a power outage and they had PoE problems did I realize there was firmware to upgrade. I upgraded the firmware at that time. The problem still persists, I still have PoE ports that should be active but are dead. The question becomes is the root cause of the problem the PoE firmware upgrade?

 

I do have this piece of code of every switch. We have had the lldp interface all for years. We added the lldp-med interface all when we added our IP phones

 

protocols {
     igmp-snooping {
     vlan all;
     }
     rstp;
     lldp {
     interface all;
     }
     lldp-med {
     interface all;
     }
}

Hello Henderson,

 

I know by experience the 4200 is quite stable regarding POE, if updating the firmware did nothing I would say there is a good chance you are facing a HW problem.

I did see once in my TAC life disabling lldp and leaving only lldp-med solved an issue other than that your next step is probably opening a case if that is a posibility. 

Henderson,

 

can you run this from shell and show the outputs:

 

root@EX4200:RE:0% lcdd 0 chassism

Chassism<0>#poe get system status

 

then we can check one of the faulty ports :

 

for POE the port ge-0/0/11 is port 11:

 

chassism<0>#poe get port 11 status​

As per PD69000 Serial Comm Protocol, version 6.6​

Actual Port Status (table 5).​

Port Enable = 1​

Latch = 0xc0​

Detection failure sticky count.​

Class = 0​

Port is on: Valid Resistor detected.​

that might give us more information 

 

 

 

 

 

Here is the output from those two commands, not sure how to interpret them. ge-1/0/25 is a port that has a wifi AP connected but is not getting PoE

 

chassism<0>#poe get system status
CPU Status 1 = 0x0 ; normal.
CPU Status 2 = 0x2; less than 8 devices.
Factory default = 0x0 ; factory default param not set.
GIE = 0x0 ; GIE Normal.
Private Label = 0x0
User Byte = 0xff
Device Fail Status = (0xf0), all devices OK.
Temperature disconnect = (0x0), all devices temperatures ok.
Temperature Alarm = (0x0), all devices temperatures under limit.
Interrupt Register = 0x7

 

 

chassism<0>#poe get port ge-1/0/25 status
As per PD69000 Serial Comm Protocol, version 6.6
Actual Port Status (table 5).
Port Enable = 1
Latch = 0x0
Class = 2
Port is on: Valid Resistor detected.

well it is basically saying there is no problem on that port it is negotiated as class 2 and a valid resistor is detected it should pass power, but it doesnt, in other examples you would see another code on the last line. 

We opened a ticket today to have the switch replaced. Once we have it replaced, I will report back

 great! hope you get the problem solved

The mystery deepens. We received a replacement switch from Juniper, installed it and the same behavior. The switch has Ruckus APs patched in on ports 0-31, ports 32-47 are empty. The APs on ports 0-24 are getting PoE power, the APs on ports 25-31 are not. This is from the Juniper documentation:

 

"The power budget calculation for class 0 and class 4 powered devices, whose power allocation is based solely upon class, is based on actual power consumption. These values are displayed in the Power consumption field in the output of the show poe interface CLI command. The power budget is the sum of the power consumption values for all PoE-enabled interfaces."

 

When we issue the "show poe interface" command it shows these APs as class 4 devices. The maximum power column shows 25.0w(L) where the L means LLDP negotiated this power. The power consumption columns show actual power usage which is between 5 and 5.6w.

 

My latest theory is the switch is not calculating the power budget based on actual power consumption like it should but rather maximum power consumption. PoE+ can deliver up to 30w per port

24 ports delivering power x 30w per port = 720w

 

The total power budget for the switch is 740w. If one more port were powered it would jump over the power budget. If the switch is not calculating the power budget correctly, this is a software issue. I will do some digging along these lines. The other way to handle this is to change poe from "class" to "static" and set max power on each port to 15w (740w total divided by 48 ports = 15.4w per port). Setting it statically to 15w per port should be able to power up all APs easily since they only draw a bit more than 5w each

As I mentioned in an earlier post, we have these WiFi APs patched into this switch for nearly 3 years. During that time we have upgraded Junos on this stack 4 or 5 times meaning a reboot. There was never a problem to now. The only change is we are nowing using lldp-med since installing a new phone system in the past month. We have always had lldp enabled but just recently added lldp-med. I know that lldp and lldp-med has capabilities around PoE. I wondering if the addition of lldp-med somehow changes the PoE parameters for this switch and now the switch thinks its PoE power budget is exhausted?

@HendersonD - in your message #29 you stated:

 

"My latest theory is the switch is not calculating the power budget based on actual power consumption like it should but rather maximum power consumption. PoE+ can deliver up to 30w per port

24 ports delivering power x 30w per port = 720w"

 

I think I stated this in my message #8.

 

Anyway, using static configuration should be a 'work-around', but does not answer the question of why.  Have you and can you capture the LLDP-MED communication between the Ruckus AP and the EX switch?  Besides enabling LLDP-MED as a change, has there been any changes on the Ruckus side, like AP/Controller SW etc.?

 

My thought is that even though LLDP-MED should be working, it is maybe not working properly, and therefore switch is defaulting to Class operation.  Under Class operation, the switch should (for Class 4 device) reserve 30W per port, which means for 24 ports 720W total or no other Class 4 device power.  For a switch that is not-configured as Static, that is 24 ports max operation, if port 25 is set to static with 15.4 W (or anything less than 20W) I would assume the 25th port then provides power, yes?

 

What is your TAC Case for this power issue, not for just the RMA?

 

I am wondeing if your situation is specific to these specific Ruckus APs and their SW?  I have not heard of anyone else having issues with EX4200-48PX and for sure know that EX4200-48PX can provide power to 24+ ports with Aruba APs - tested this myself.

When I run this command I get nothing

show log messages | grep poe_get_psu

 

root@HS102B> show configuration poe | display set
set poe interface all

Hi Henderson,

 

Here are my 2c: 

a) With the workaround (static POE) removed, please capture:

show poe interface <interface_name>
monitor traffic interface ge-0/0/0 extensive write-file /var/tmp/lldp.pcap

 

b) FTP the lldp.pcap from the switch to your PC/MAC. Use Wireshark to open the sniffer.  Check the power class in the POE TLV sent by the EX to the AP.  It should match what's discovered on the POE interface.

 

c) If there's a mismatch, please check if you're running one of the JTAC recommended releases, if not please upgrade to it:

https://kb.juniper.net/InfoCenter/index?page=content&id=KB21476&actp=METADATA

 

There's a chance we're hitting this bug: https://prsearch.juniper.net/InfoCenter/index?page=prcontent&id=PR1296547

 

Hope this helps.

 

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

The JTAC recommended versions for the EX4200 are Junos 12.3R12-S12 / 15.1R7

I am running Junos 15.1R7.9

Ok, did you check the POE class with static removed on a problem port.

 

Hope this helps.

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

 

I did not use the method yet that you described. When I use "shoe poe interface" the Ruckus APs are shown as class 4 devices. They were shown this way when the PoE was set to "class" and it did not change when I changed to "static". I did reboot the entire stack earlier this morning so this is accurate information in static mode. Under the Max Power column it is now showing 15.4w where before it was showing 25.0w(L) where L indicates that lldp/lldp-med negotiation set the power.

 

In class mode this should work according to this taken directly from Juniper's documentation.

"The power budget calculation for class 0 and class 4 powered devices, whose power allocation is based solely upon class, is based on actual power consumption. These values are displayed in the Power consumption field in the output of the show poe interface CLI command. The power budget is the sum of the power consumption values for all PoE-enabled interfaces."

 

Would changing it back to class and using your method to actually see what information is being set during negotiation help? If the Juniper switch in class mode sees these APs as class 4, shouldn't they just work?

 

 

 

Yes they should, but the procedure I showed confirms if the switch actually responds to Class 4 or not. If it's not responding, that's a match of the mentioned bug and you should go to JTAC with this information.

Hope this helps.

Regards,
-r.

--------------------------------------------------

If this solves your problem, please mark this post as "Accepted Solution."
Kudos are always appreciated :).

Juniper Internal

I have my local Juniper SE and my local Ruckus SE looking into this. I will let you know what I hear. In the end I may take all of my switch stacks and change them from PoE class to PoE static

Actually my bad all along, sorry!  When using LLDP-MED the Power Budget calculation is based upon LLDP-MED value (in your case 25W) plus a 'power-guard' for potential cable loss (@100M).  So for 25W from LLDP-MED TLV, calculation (plus ~15% for power-guard) is around 30W.  This is independant of the actual power draw.  In this case for 740W supply, max number of POE interfaces is 24 (720W total using LLDP-MED value for calculation).  From docs:

 

When the management option for PoE is set to class and LLDP is enabled (both are default settings), LLDP power negotiation is enabled by default. If you disable LLDP power negotiation or the powered device does not support it, the switch uses the class of the powered device to determine the maximum power for interfaces.

 

To change this behavior, the recommendation would be to use Static configuration instead, which you already know works differently.

 

Sorry it took me so long to remember this properly, and I am not sure Juniper documentation is 100% clear about this.

I cannot find anywhere in Juniper's documention that for LLDP-MED the power budget is the negotiated value (25.0w in my case) plus a cushion. Here is the guide for PoE power on EX switches

https://www.juniper.net/documentation/en_US/junos/topics/concept/poe-overview.html

 

Two sections that might apply. The first section below says the PoE budget should be based on actual power consumption

"The power budget calculation for class 0 and class 4 powered devices, whose power allocation is based solely upon class, is based on actual power consumption. These values are displayed in the Power consumption field in the output of the show poe interface CLI command. The power budget is the sum of the power consumption values for all PoE-enabled interfaces."

 

This is from the LLDP-MED section of the document I linked to. I think the key section is this "The PoE controller allocates to an interface only the power currently required by the connected powered device". These Ruckus APs only need 5-6w of power, they do not need 25.0w of power. My best guess is the Ruckus APs are telling (incorrectly) the Juniper switch, set aside 25.0w for me.

"Link Layer Discovery Protocol (LLDP) power negotiation enables the PoE controller to dynamically allocate power to LLDP-enabled powered devices based on their power needs. The PoE controller allocates to an interface only the power currently required by the connected powered device, plus some additional power guard to accommodate cable length. This additional allocated power is approximately 15 percent of the requested value and it can allocate the power in small increments."

My bad, a little farther down in the document is this. The "power budget calculation is based on requested and negotiated power value". So the question is, why is the Ruckus AP telling the Juniper switch to set aside 25.0w when it only draws about 5-6w?

 

"In LLDP power negotiation mode, the power budget calculation is based on requested and negotiated power value, plus approximately 15 percent more. The maximum power for an interface is displayed in the Max power field in the output of the show poe interface CLI command. The power budget is the sum of the maximum power values for all PoE-enabled interfaces. In the case of LLDP power-negotiated devices, the total power consumption shown in the output of show poe controller is always less than the actual POE power budget."

Because Ruckus (and others) could pull that much wattage under other conditions.  BTW, Aruba (for 100% sure) and other WiFi vendors (including Juniper MIST) all do the same.  Under some conditions, maybe not now but in future, AP could actually draw 25W (30W with power guard).  This is per the spec.

 

If you set static to say 15.4W or something smaller, there could be a condition where AP does not come up if it requires greater than 15.4W.  Make sense??

 

HTH

The EX4200-PX switch can deliver up to 740w of power. We did initially set all ports to static for PoE at 15.4 watts. Later that day I recevied a call from my facilities manager asking why a bunch of outdoor security cameras were down. Apparently they draw 16w in good weather and up to 25w in bad weather. Setting every port to 15.4w would not cut it. To solve our problem we turned off lldp power negotiation using this command

set protocols lldp interface all power-negotiation disable

After rebooting the stack, everything worked fine. Wifi APs, security cameras, and IP phones are all gettting power.

 

I did have a discussion with Ruckus support on this and they agreed that the way the power negotiation and power budgets are handled could use some work. I wish the power budget was just actual consumption. In my case our switches can deliver 740w of power and the largest power drawn we have seen is 170w.

Thank you for this post! We currently have Ruckus and Juniper as well, and I have noticed inconsistencies in regards to power. In some cases the device shows it negiotiated through LLDP, but most of the time it doesn't. I will try this test and see if it fixes our issues as well, as some devices seem to not work all the time and some of our APs have dropped out randomly.

 

So for this fix, you basically just undid all of the settings mentioned previously, and from the default settings, just issued this one command correct?

That is correct. In the end this is what the two sections of pertinent code look like

This section has the piece that turns off lldp power negotiation

protocols {
   igmp-snooping {
      vlan all;
   }
   rstp;
   lldp {
      interface all {
         power-negotiation {
            disable;
         }
      }
   }
   lldp-med {
      interface all;
   }
}

 

This section turns on PoE for all ports

poe {

    interface all;

}

EX4200 has the capability to support as much as 740 Watts for POE, but what happens when the actual power source delivered to the switch is less than 740W?

In this case the although the output of the show poe controller command may so total available POE power of 740, in reality the switch might not be able to provide power of 740 Watt. So in this case it might be worth checking if the circuit breaker used to connect the switches is actually capable of providing a power equal to or greate than 740W.

Also its important to remind oneself that the total power load of a circuit breaker can be divided into multiple electrical outlets. If for example a 15Amp breaker, able to produce 1800 Watts of power with 120V input is diving among 3 electrical outlet and if it so happens that 2 of the outlets use most of the 1800 watt power, the last outlet will be left with very less wattage to provide.
Just something to keep in mind.

When performing the poe check, we get an error:

 

chassism<0>#poe get system status
CPU Status 1 = 0x0 ; normal.
CPU Status 2 = 0x3; PoE Controller memory error.
Factory default = 0x0 ; factory default param not set.
GIE = 0x0 ; GIE Normal.
Private Label = 0x0
User Byte = 0xff
Device Fail Status = (0xf0), all devices OK.
Temperature disconnect = (0x0), all devices temperatures ok.
Temperature Alarm = (0x0), all devices temperatures under limit.
Interrupt Register = 0x7

 

What does "PoE Controller memory error" mean? This is a Junos EX3300 running 12.3R12-S12 for reference.