Comware

We have troubles with our HP 9304 routing switch. For the most time it has 1% cpu utilization, like this:
telnet@hp0#sh cpu
13 percent busy, from 61 sec ago
1   sec avg:  1 percent busy
5   sec avg:  1 percent busy
60  sec avg: 11 percent busy
300 sec avg: 25 percent busy

But sometimes (approx. 1 time per minute) it has 99% CPU load like this:
telnet@hp0#sh cpu
67 percent busy, from 8 sec ago
1   sec avg: 95 percent busy
5   sec avg: 80 percent busy
60  sec avg: 20 percent busy
300 sec avg: 27 percent busy

As a result we have 20-30 average CPU load, command 'sh proc cpu' shows nothing intresting, CPU utilization of all processes less then 0.4% for each process.
This CPU Utilization fluctuations looks like periodical process and period decseases with increasing common traffic though the switch.

Does anybody know how to find the reason of such peaks in CPU load and how to solve this problem ?

Hi

Some interesting traffic should be processed by the CPU before forwarding.
Example: Source-Based routing, traffic controlled by ACLs, DSCP Mapping, NAT, 802.1x traffic .....

Since this utilization is not happening most of the time, so you may consider that one of the interesting traffic happens in sometimes that makes the CPU busy for a while.

The most common traffic that occur in most networks and make the CPU busy is Broadcast, Spanning-Tree calculation, so check for consistent MAC address clearing or spanning-tree instability.

Good Luck !!!

HI,

Another thing to look out for on the broadcast aspect of networking, is trunks.

If you are using LACP to create dynamic trunks and the link is lost on one of the ports, a broadcast is created while the trunk fluctuates.

I would check the logs and the config of LACP.

Try disabling LACP on all ports that do not require it.

and try to use Static trunks over Dynamic.

I hope this helps.

Jase