C3850 high CPU caused by IOSD

[Marius Gunnerud]

We have a 3850 switch running version 03.07.04E that is reporting high CPU caused by "EPC WS Pkt Send p"  I have been trying to figure out what could be causing the Embedded Packet Capture to cause high CPU but am not having any luck.  Anyone seen this behavior before?

GM-TI14RO131#sh proc cpu det process iosd | ex 0.0
Core 0: CPU utilization for five seconds: 6%; one minute: 11%; five minutes: 32%
Core 1: CPU utilization for five seconds: 99%; one minute: 94%; five minutes: 47%
Core 2: CPU utilization for five seconds: 18%; one minute: 5%; five minutes: 31%
Core 3: CPU utilization for five seconds: 6%; one minute: 3%; five minutes: 3%
PID T C TID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
(%) (%) (%)
12306 L 3097934 236130264 865 24.6 24.8 24.8 34816 iosd
12306 L 1 12306 3962820 198973711 0 24.5 24.5 24.6 34816 iosd
12306 L 2 13637 2125715 165584366 0 0.20 0.18 0.19 34816 iosd.fastpath
7 I 1785523 2017211 0 0.33 0.33 0.22 0 Check heaps
16 I 4097891 53012362 0 99.1 98.6 98.9 0 EPC WS Pkt Send p

[Marius Gunnerud]

This seems to have been a bug.  Although I have not found an exact bug description match, this comes fairly close CSCux39490

After a reload of the switches the problem when away...for now :-)

[Julio E. Moisa]

Hi Marius,

It could be related to that bug, also try to open a ticket with the TAC in order to verify what triggered that high CPU.

:-)

[Florin Barhala]

Hi Marius,

I found this bug ID https://bst.cloudapps.cisco.com/bugsearch/bug/CSCvc08452/?rfs=iqvred
Since 2017 did you experience this behavior? What SW version are you using now?

I am on the same boat for the past couple weeks: I had to tshoot some QoS incident, enabled EPC no matter what SWs have become close to unresponsive.

Thanks!