ExtremeSwitching (EXOS)

 View Only

 SIP-T21(P) E2 YEALINK IP PHONES IN THE NETWORK CREATE CONGESTION ON PORTS

abhijit_roy's profile image
abhijit_roy posted 04-29-2022 04:57
Hi All,

      I am facing an issue in last couple of months, I have 4 no's of X440G2-48p-10G4 access switches in the network, one of the switch in the network got connected with above mentioned IP Phones like as follows due to scarcity of the physical port

 User End Port Connected to Internet port of IP Phone, and P.C Port of IP Phone connected to Desktops, I have observed that on those particular ports where IP Phones are connected facing port congestion issues,  and counters are incremental, and fluctuation in voice calls, for that for testing I have bypass the IP Phone and connect one to one like P.C to Switch and IP Phone to Switch, on that scenario I have observed port congestion in ports connected to IP Phone and there is no congestion in P.C Port, all the IP phones are apparently new.

So I was wondering if I have to enable any specific Protocol like lldp or anything else to avoid this scenario for access switches, and also please suggest about the recommended practice to connect IP Phones in the network.
Gabriel_G's profile image
Gabriel_G

Hello,

I would be suspicious of pause-frames from the VOIP phone causing this issue. Pause-frames are messages sent by an end device to tell a link-neighbor to slow down. This can in turn cause congestion for that port, especially if there is a work-station behind the VOIP phone.

To check for incoming pause-frames, check the command:

show port flow-control

In EXOS, you can tell the switch to ignore pause-frames with the following command:
disable flow-control rx-pause port <#>

Generally, ignoring pause-frames will not negatively impact the VOIP phone, but there may be a limitation of the VOIP phone as to how fast it can switch frames to the workstation.

Hope that helps!

abhijit_roy's profile image
abhijit_roy
Hi ,

Thanks for your suggestion, also kindly suggest what will  be the  Extreme recommended standard practice to set up network on this kind of situation, and how to detect if there is any error on  end point devices or switch port vice versa on such case?

Thanks
Gabriel_G's profile image
Gabriel_G
Hello,

I'm not sure what you're asking specifically; their is no official recommendation to be made.

However, note that you can check for layer 1 errors on the switch port via the command 'show port <#> rxerrors'.

Hope that helps!
abhijit_roy's profile image
abhijit_roy
Hi Gabriel,

Following is my concerns in detail for this case

This type of connections of IP Phones has been enabled in my network from last couple of months, and following are consequences of such(I am suspicious of that)

a) Some P.C internet connections/lan got hang we have to reboot the p.c to come back to the network.
b) During conversation of users between IP phones sometimes voice got distorted, we have to disconnect the call and reconnect, it seems like network fluctuations.
c) As I have said earlier that for testing I have made direct connection to IP Phones and Desktops from switch, on such scenario also port congestion found in IP phones and there is no congestion on P.C, so my doubt is I am doing any mistake in process to connect any IP Phone in network, or is this IP Phones have any problem(Though supplier are said all IP Phones are quite new).
d) And this IP Phones are 10/100, and current x440 switch is 10/100/1000 with auto on, for that is there any issue.
e) Also P.C and IP Phones are in same sub vlan, for that also there may be any issue?

Overall my concern is to get permanent solution of this problem, as users are facing problems frequently as I have described above, Kindly suggest

Thanks again for your cooperation and support.
Gabriel_G's profile image
Gabriel_G

Hello,

I'd be concerned with the phone only supporting 10/100 Mbps connections as having a workstation behind the phone will likely congest that link.

I'd check the following from the switch:
1) Are other wired clients having issue or just IP Phone/Workstation ports? Checking if your wired network is stable in general.
2) show port <#> rxerrors
any errors here indicate a layer 1 issue (bad port/cable)
3) show port <#> flow-control
if you see pause-frames, that could cause congestion. you can test with disabling pause-frames with this command:
disable flow-control rx-pause port <#>
4) show port <#> congestion
If you see congestion after disabling pause-frames, you are over-utilizing the link to the IP phone which may happen if the phone only supports 10/100 Mbps connections.

I'd recommend creating a GTAC case if you're still having issues.

Hope that helps!

abhijit_roy's profile image
abhijit_roy
Hi,

  I have observed in detail, as per your suggestion, following is the details
1. There is some congestion in some workstation ports in network, but there is no major problem.
2. There is no rxerror.
3. yes in show ports flow-control there is  huge congestion in ip phones port.
4. yes after disabling rx pause also I am facing port congestion for IP Phones.
Gabriel_G's profile image
Gabriel_G
Hello,

It sounds like the links are simply being over-utilized causing port congestion. Having a workstation connection and also an IP phone may be too much traffic for a 100M link. You can check 'show port <#> utilization' and press [spacebar] to see how much you're utilizing those ports during congestion.
abhijit_roy's profile image
abhijit_roy
Hi,

  Sorry for delayed response, here is following statistics of port utilization

Link Speed  Rx % bandwidth  Peak Rx % bandwidth  Tx % bandwidth Peak Tx bandwidth
100                 0.00                        2.38                                     0.08                               4.74


Rx Pkts/sec    Peak Rx pkts/sec     Tx Pkts/sec    Peak Tx pkts/sec
        0                         286                          16                         498

Rx bytes/sec    Peak Rx bytes/sec    Tx bytes/sec      Peak Tx bytes/sec
       74                       427471                      3821                      582670


It will be very helpful if you can suggest any alternative solutions of this problem.
Gabriel_G's profile image
Gabriel_G

Hello,

Could you confirm how long you observed the utilization and if the port congestion was happening at that time?

Note that the port utilization statistics are 5s averages from when you start looking at the output of the command. If you collected this output when there was no congestion, it may not show high utilization. You can also have congestion with low utilization due to bursty traffic.

Brandon Clay's profile image
Brandon Clay
Hi,

Were the utilization counters cleared prior to checking them? Or was there congestion present while checking them?

If congestion was not present and the counters had been cleared, any possible overutilization may not be apparent from the counters.