WS-C3750X-48PF-S CPU Utilization

[engmkhalifa]

First of all I purchased the switches with part no. WS-C3750X-48P with the following IOS c3750e-universalk9-mz.150-1.SE1 & IOS feature license to work as ipservice.

Symptoms:

1)      - When running the switches in stack mode (connecting the booth switch with full-duplex stack) I found the cpu utilization on each switch reached to 34% The Hulc LED Process reaches to 20%

3)      - After about 1 hour  with no connections to the switch I found the status not changed.

4)      - I tried to run each switch in stand-alone mode (without stack cables connected) & monitor the status of the cpu utilization. I found it the same problem exist.

Actions taken:

1)      - Refereeing  to the internet I found a bug ID describing the status of the mentioned symptoms (attached)

2)      -  I made a down grade for the IOS version to be  c3750e-universalk9-mz.122-58.SE2.bin

3)      -  The mentioned IOS solved the Hulc LED Process (became 1 : 3 % ) & however cpu utilization became 18% with no links or traffic connected to the switches.

4)     - I guess that 18 % utilization is too much for a free switch (refer to the show run file) as I did not configure any advanced commands.

please advise with a suitable solution for the mentioned case ASAP

the required logs are attached

[Arun Nair]

Just shut down the non-requried ports(many i believe ). I have faced the same issue.

Cheers

Arun

[glen.grant]

  Thats not too high . People get too hung up on cpu that is above 20% .  The cpu generally has little to do with traffic switching these days which is handled in hardware. A 4500 will run 30-40% with nothing attched , this is normal also.

[Leo Laohoo]

There are CPU bugs that are currently haunting 12.2(58)SE and 15.0(1)SE.  I'm testing 15.0(1)SE2 and it's alot better.  How about trying that?

[engmkhalifa]

i already tried this version before without effect.....

[NormMuelleman · Accepted solution]

Just my nickel's worth (a penny is worthless anymore)...

I worked at a site where we had Cisco 7206s, 3750's, etc. We used CA's Spectrum to monitor the CPU's.

And at my current site, we are running multiple 3750's in Layer 3 mode...

We don't get notifications unless the CPU goes above 50%, or bandwidth goes above 80%.

I understand that nothing is attached. But you'll have to compare a bare-bones device vs what you're running to see if there's a cause for concern.

I just looked at one of our 3750's. Granted, it's in the middle of the night here, but it's sitting at 13%.

I looked at another one; it's the master of a 5-stacked switch. It's sitting at 16%. It's average runs between 15-40%.

Just trying to give you perspective

[ashish shirkar]

This might help 

https://supportforums.cisco.com/discussion/11145246/high-cpu-hulk-led-process