Unexpected Link Speed Drop on Cisco EWC Interface

[Radu96069]

Hello!

I’m facing the following issue:

Recently, I noticed that the Cisco EWC gigabit interface on a C9130 access point is negotiating at only 1000 Mbps. This surprised me because I’m certain it was running at 5000 Mbps when I initially connected it to the switch (around three years ago).

Interestingly, when I check the same interface directly on the AP, it reports 5000 Mbps full duplex. (The EWC and the switch both report 1000 Mbps).

The switch ports and cables (Cat 5e and Cat 6a) appear to be fine. I tested them with a C1972 AP, which successfully negotiated a 2.5 Gbps full-duplex connection. As mentioned, the C9130 EWC used to work at 5 Gbps previously.

Does anyone have an idea why this might be happening and how to fix it? I suspect it could be a bug, but there might also be a configuration setting limiting the EWC's speed.

Thanks!
Radu

[marce1000]

     - When such events happen, check the logs on the connecting-switch and the EWC-AP , or (preferred) : for both use a central syslog server, so that you can correlate events,

  M.

[Radu96069]

Thanks!
There is nothing in the logs (I've restarted the EWC-AP) - probably I need to change it to debug level.

[Scott Fella]

So need to clarify things here: CruzBarrera

The switch ports and cables (Cat 5e and Cat 6a) appear to be fine.
Is this with a cable tester end to end or just patch to switch, or just because you connected another ap?

I tested them with a C1972 AP, which successfully negotiated a 2.5 Gbps full-duplex connection.
Are you using the same port? Because 2.5Gbps is aTwoGigabitEthernet mGig port vs a 10G (TenGigabitEthernet) mGig port that can negotiate to 5G. The 9172 connected at 2.5G not 5G like you mentioned your 9130 did previously?

As mentioned, the C9130 EWC used to work at 5 Gbps previously.
So your ap is connected to a 10G port and no one has moved that ap to a different port?

You should also post some screen shots of what you are seeing?

[Radu96069]

Thanks, Scott!

1) Same cables I used when I installed it for the first time and it worked (5e). I checked with the other ap and it negotiate 2.5 Gbps. I check with a short 6a patch both aps on multiple ports and on all of them the C1972 worked at 2.5 and the 9130 EWC only at 1 Gbps. So, I assume is not the cable. No cable taster.

2) Yes, I used the same port, and I tried others. Same results. The switch can deliver PoE ++ on all ports and I tried the 2.5 max and the 2.5, 5, 10 ports. Once again it was on the same port I tested some years ago and it worked.

3) no. It is the same mGig port. It's placed in a home office and there are no other devices on that specific switch.

What exactly do you want to see?
A show interface gig 0 on EWC show 1000000 kbs on the AP shows 5000 Mbps. See below

- EWC -

MTU 1500 bytes, BW 1000000 Kbit/sec, DLY 10 usec,
reliability 255/255, txload 1/255, rxload 1/255
Encapsulation ARPA, loopback not set
Keepalive set (10 sec)
Unknown, Unknown, link type is auto, media type is unknown media type
output flow-control is unsupported, input flow-control is unsupported
ARP type: ARPA, ARP Timeout 04:00:00
Last input 00:00:00, output 00:00:00, output hang never
Last clearing of "show interface" counters never
Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0

- AP -
UP BROADCAST RUNNING PROMISC MULTICAST MTU:1500 Metric:1
collisions:0 txqueuelen:80
Base address:0x7000
full Duplex, 5000 Mb/s

5 minute input rate 733947 bits/sec, 77 packets/sec
5 minute output rate 16919 bits/sec, 10 packets/sec

[Rich R]

@Radu96069 Gig 0 on EWC is a Virtual Interface in a container/virtual machine - as discussed on other threads here.
That's also why it says: "Unknown, Unknown, link type is auto, media type is unknown media type"
So the "speed" on EWC Gig0 is meaningless - it simply reflects the interface name!
There's a whole lot of strange/anomalous behaviour in the EWC IOS-XE as a result of that virtual interface - things that don't work the way you'd usually expect on a real interface - commands that have no effect (even though IOS-XE still allows them to be entered sometimes, just because the command has not been removed).

What you're seeing on the AP apparently confirms it is running at 5Gbps and you can also check the switch side of the link to confirm that too with show int and show int status, although you say "The EWC and the switch both report 1000 Mbps"

Let's see the show int and show int status from the switch?

That said - there are a number of bugs in the APs and Cat 9K switches with mgig, so the code version (on both sides) matters, but you have not mentioned what versions you're running on the EWC or the switch or what model the switch is.

[Radu96069]

Hi @Rich R 

Thanks for the information.

The switch is not Cisco. Is a Netgear MS510TXUP running the latest software version (1.1.0.9)

NETGEAR# show interfaces XMultiGigabitEthernet 5 status
Port  Name                 Status      Vlan  Duplex  Speed    Type
xmg5                       connected   1     a-full  a-1000M  Copper
NETGEAR# show interfaces XMultiGigabitEthernet 6 status
Port  Name                 Status      Vlan  Duplex  Speed    Type
xmg6                       connected   1     a-full  a-2500M  Copper

XMultiGigabitEthernet5 is up
  Hardware is Ten Gigabit Ethernet
  Auto-duplex, Auto-speed, media type is Copper
  flow-control is off
  back-pressure is enabled
     3937030 packets input, 2753470702 bytes, 0 throttles
     Received 38441 broadcasts (115266 multicasts)
     0 runts, 0 giants, 0 throttles
     0 input errors, 0 CRC, 0 frame
     115266 multicast, 0 pause input
     8 input packets with dribble condition detected
     14404766 packets output, 404148573 bytes, 0 underrun
     0 output errors, 0 collisions
     0 babbles, 0 late collision, 0 deferred
     0 PAUSE output

NETGEAR# show interfaces XMultiGigabitEthernet 6
XMultiGigabitEthernet6 is up
  Hardware is Ten Gigabit Ethernet
  Auto-duplex, Auto-speed, media type is Copper
  flow-control is off
  back-pressure is enabled
     3796091 packets input, 3882548572 bytes, 0 throttles
     Received 10787 broadcasts (16144 multicasts)
     0 runts, 0 giants, 0 throttles
     0 input errors, 0 CRC, 0 frame
     16144 multicast, 0 pause input
     4082116 input packets with dribble condition detected
     8397455 packets output, 1713055219 bytes, 0 underrun
     0 output errors, 0 collisions
     0 babbles, 0 late collision, 0 deferred
     0 PAUSE output

The EWC is on port 5.
I've tested the throughput also. When I installed the switch, I could easily get 1.2 Gbps (Dell Wi-Fi 6 160 Mhz channel. Today I barely get 550 Mbps same as before moving the AP on it. The C9172 performs as expected (1.2 Gbps on 6ghz with an iPhone)

Thanks!

[Scott Fella]

Wireless is half-duplex, so to get that kind of speed on wireless is really rare.  When you test throughput, you should test with iPerf or Openspeedtest on a machine connected at 10G wired.  Then test iPerf/Openspeedtest on another wired machine that supports 10G and test again on a 1G connection.  Get at least a baseline of the wired because that test would be full-duplex.  When you test on wireless, you should at max maybe get half of that, but be realistic on what the wireless card can do.  Do not test to the internet, but locally in your LAN.

There are a lot of things that affect the final throughput results, interference, noise, channel width, signal, etc. As far as what the ap and switch negotiate to, you need to troubleshoot that more. Look at what the ap is negotiated to and the switch not the EWC. What get's me is that you thought it negotiated 3 years ago at 5G and a new 9170 only negotiated at 2.5G.  So again, that is something you will need to figure out, maybe hit up the Netgear forum also.

[Radu96069]

Thanks, Sott!
The iperf server is local and is connected to the switch using a 10Gbps fiber port. I know what to expect in terms of Wi-Fi performance.
9170 cannot negotiate higher because it has a 2.5 gig only port (1x100M/1000M/2.5G Multigigabit Ethernet (RJ-45) Uplink)
I will have a look on Netgear forums, but my assumption was that if other devices work than the problem is the EWC ap.

Thanks again!

[Rich R]

And what software version is the EWC running?
Did you try rolling back to the software version which you thought it negotiated higher speed? (backup your config first of course)
Interesting that the switch and AP think the speed is different - could be a cosmetic bug on the AP.

[Radu96069]

17.12.4.
No, as I didn't have time for this and also, I do not remember what version it was. I could probably figure it out, but I do not think it worth the effort. It doesn't impact the client's internet performance and anyway I plan to move the AP to a WLC-CL for testing purposes, and I will see then if the lightweight ap has the same problem.
Anyway, if debugging this helps anyone let me know and I could invest more time in in.

Thanks for your support!
Radu