3850 switch high CPU issue

[mohamed hassan]

hi all ,

im facing an issue with a 3850 stacked switches,

NGWC l2m process is consuming CPU cycles,

we are not even using this switch as wireless controller 

any ideas ?

thanks !

[InayathUlla Sharieff]

Hi,

NGWC L2M is responsible for IPv6 multicast traffic, without snooping the traffic would be
flooded and forwarded in software resulting in high CPU.
When MLD snooping is enabled, a per-VLAN IPv6 multicast address table is constructed in
software and hardware.
The switch then performs IPv6 multicast-address based bridging in hardware, avoiding to be
processed by software.

Resolution Summary :
conf t
ipv6 mld snooping
end
copy run start

HTH

Regards

Inayath

**Please do not forget to rate the post if this is helpfull*******

[mike-maier]

This resolved the problem with our 3850 stack that has been on and off for sometime.  Thanks.

[ROBERT PASKE]

Seeing the same issue on a 3850 stack.  Resolution resolved our issue.

n164-3850-DHS-SP-1stFL-1#show process cpu detailed process iosd sort | exclude 0.0

Core 0: CPU utilization for five seconds: 38%; one minute: 37%; five minutes: 29%

Core 1: CPU utilization for five seconds: 29%; one minute: 63%; five minutes: 58%

Core 2: CPU utilization for five seconds: 32%; one minute: 43%; five minutes: 44%

Core 3: CPU utilization for five seconds: 97%; one minute: 57%; five minutes: 69%

PID    T C  TID    Runtime(ms) Invoked uSecs  5Sec      1Min     5Min     TTY   Process

                                               (%)       (%)      (%)          

13966  L           2005560     4232774 289    24.44     24.34   24.30   34818 iosd

13966  L 0  13966  274971      3492131 0      24.00     23.91   23.88   34818 iosd

13966  L 1  15997  2702570     4519213 0      0.44      0.38    0.37    34818 iosd.fastpath

156    I           1194280     688099  0      94.88     94.77   94.77   0       NGWC L2M

234    I           4229300     1802961 0      1.33      1.44    1.44    0       Spanning Tree

19     I           1415978     1989782 0      0.66      0.22    0.22    0       CMI IOSd task

93     I           3950035     1220958 0      0.55      0.44    0.44    0       PLATFORM_MGR SPI in

211    I           338384      6621892 0      0.55      0.44    0.44    0       CDP Protocol

65     I           3997452     5933244 0      0.44      0.44    0.33    0       Net Background

195    I           3971998     2511609 0      0.33      0.11    0.11    0       Inline Power

[varun.sheela@wipro.com]

Try this to reduce the CPU

3850#conf t

Enter configuration commands, one per line. End with CNTL/Z.

3850(config)#ipv6 mld snooping

3850(config)#end

https://www.cisco.com/c/en/us/support/docs/switches/catalyst-6500-series-switches/200126-High-CPU-Utilization-on-Catalyst-Switche.pdf

Regards Mihir.

[Mattias Lindgren]

This resolved an issue for us on a 3650 stack as well, running 03.06.04.E

[Feds]

Same here! Fixed for 3.3.0 (I know... 2020 and still running that version).
Thanks Inayath

[Luis Rodolfo Campos Garcia]

I saw this same behavior just 2 days ago in a 3850 Stack running  16.3.5b (actually cisco's recommended) code

Fixed the issue with the solution provided above lowering the CPU usage from 99% to 79% now the process consuming the majority of CPU is ipv6 mdl snooping

Even after enabling ipv6 mld snooping, had to disable spanning-tree loopguard because of false positive even with all copper ports except 1 in shutdown state... loopguard kicks in and in fact traffic disruption occurs.

TAC time