High CPU Utilisation

[haidar_alm]

Hello,

Users are reporting some lag issues when connecting to the network, upon running the sh processes command I noticed that Marvell wk-a Pow process is running at a 0.97%.

I've tried to look online and on the cisco site as to what this process does without any luck!

Can someone please advice or provide some info on this process please?

Many thanks.

Below is the output:

sh processes cpu sorted 5min

CPU utilization for five seconds: 7%/0%; one minute: 7%; five minutes: 7%

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

184    29335958   5201982       5639  0.47%  0.95%  0.97%   0 Marvell wk-a Pow

105     6357851  56712192        112  0.31%  0.29%  0.29%   0 Hulc LED Process

   7     2608224   7260941        359  0.31%  0.24%  0.25%   0 ARP Input       

158     3281714   9012165        364  0.31%  0.20%  0.17%   0 Spanning Tree   

125     2427682   2368348       1025  0.31%  0.16%  0.16%   0 PI MATM Aging Pr

113     4561691    476450       9574  0.15%  0.16%  0.15%   0 HQM Stack Proces

114     3620140   1902788       1902  0.15%  0.10%  0.11%   0 HRPC qos request

   4     3521340    322493      10919  0.31%  0.12%  0.11%   0 Check heaps     

  76     2274131   4967334        457  0.15%  0.09%  0.10%   0 hpm counter proc

151     1169947   2917181        401  0.00%  0.04%  0.05%   0 CDP Protocol    

  77     1207975   5163700        233  0.00%  0.04%  0.04%   0 HRPC pm-counters

156     3697500  33067692        111  0.00%  0.01%  0.03%   0 IP Input        

  55      156264  70630510          2  0.15%  0.01%  0.00%   0 HLFM address lea

[Peter Paluch]

Hello,

I don't think that the lag your users are experiencing is related to a CPU, based on your current output. The CPU utilization is mere 7% which is normal.

What kind of "lags" do your users experience? Can you provide more detail? What is the device you have provided the output from? Is it a switch? If so, are the access ports configured with spanning-tree portfast? The STP is frequently the cause of "lags" when connecting to a network. In networks with RSTP, it is particularly important to have the access ports configured with this command to prevent them from being temporarily blocked in cases of topology changes.

Best regards,

Peter

[haidar_alm]

Port-fast configured, and spanning tree is on PVST and not RSTP.

Upon investigating only 3 -4 users are experiencing these issues and not all the users that connect to the same switch.

It's a 3750 48PS switch.

This doesn't sound like a network issue, could be down to old Pcs, i've rang our Desktops team to investigate. Initial log came in as all users experiencing problems, however, upon investigation, all are good except for 3-4.

Any idea what that process does by the way?

[Peter Paluch · Accepted solution]

Hello,

About the PortFast being configured - very good. Regarding the PVST, I would personally suggest moving to RSTP (RPVST+) in some future maintenance window to increase the reconvergence speed of your network in case of topology changes, or even considering MSTP if you have many VLANs.

Regarding the lag issue, I suggest having a look on the interface counters on the ports connecting to these users - check for all error counters. Perhaps there is a physical problem. In any case, the user description of "lag" should be made more precise - whether the "lag" influences all communication or only some destinations, if only increased delays or also packet losses are detectable - we need to have some evidence for the "lag".

Regarding the Marvell wk-a Pow - I don't know for sure. I can only guess from the name that it is a process that somehow deals with PoE, as the Marvell is a manufacturer of networking chips, controllers and ASICs - but this is only a very rough guess.

Best regards,

Peter

[haidar_alm]

Hi Peter,

Cheers for your input, I agree with regards to RSTP, we do have it implemented on part of our network and are trying to have it across the board depending on maintenance window as per each location.

I did not hear anything back from the user with regards to slowness, no more. We usually get a bit of a lag when accessing the CLI on the switch due to SNMP high utilization, but i've never seen this Marvell wk-a Pow or able to find any information on it!

KR

Haidar