High CPU in Catalyst 3850

[Eriks Rugelis]

A Catalyst 3850-24P runs IOS-XE 3.2.2SE is reporting what seems like excessively high rates of CPU utilitization.

There is no debugging enabled at this time.

The busiest process seems to be 'fed'

k390-cal-r-4a-0#show proc cpu sorted

Core 0: CPU utilization for five seconds: 59%; one minute: 53%;  five minutes: 55%

Core 1: CPU utilization for five seconds: 56%; one minute: 58%;  five minutes: 64%

Core 2: CPU utilization for five seconds: 45%; one minute: 54%;  five minutes: 59%

Core 3: CPU utilization for five seconds: 87%; one minute: 86%;  five minutes: 72%

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

5576   1507975     10740467 167    51.59    52.54    51.83    1088  fed               

9123   1448242     25838224 186    8.53     7.38     7.03     0     iosd              

6084   1746809     15059128 63     1.45     0.99     0.96     0     pdsd              

5577   852155      83677404 11     0.14     0.10     0.09     0     platform_mgr

What is the purpose of the 'fed' process?   Why is it so busy?   Should we be concerned?

On the 3850, are all four CPU cores treated as equal resources?   Is load on any one core significant and something to be watched for?  Which core is the most important one?

At present this switch is serving as a router for two stacks of access layer switches (a total of about 300 users.)   The only ports in use are on the C3850-NM-4-1G (one uplink, two downlinks.)

[daniel.dib]

No debugs running? There was a similar post here:

https://supportforums.cisco.com/thread/2212348

Seems this platform has some CPU load usually.

Daniel Dib
CCIE #37149

[lucky3012]

Was there ever a posted Fix for this?  I seem to be experiencing the same thing and no debugging is turned on.

Thanks

[Eriks Rugelis]

Further investigation at the time of the originally reported event revealed that forwarding of payload traffic was impaired for a most (but not all) end-users.   The solution was to reload the switch stack.

Subsequently, we have noticed the same symptoms occurring in two other stacks of 3850's.   Reloads cured them as well.

Root cause is still unknown.   Our operations team now knows this as the 'insane router' syndrome.

[lucky3012]

Thanks for the update!  When you Reloaded the stacks did you upgrade them to 3.3.0SE or leave them at 3.2.2SE?

Any changes in the reload - i.e. turn off nmps?

[Eriks Rugelis]

So far we have made no changes to software version nor to configuration.