CPU haywire on 25xx

[Paulen]

All,

I have a quite big network with a lot of ProCurve switches. Lately I have noticed that my some of my users complain about slow response.

I typically have 3 VLANs on a switch; one for management, and two for the customer. The customer has a router, thus no MAC flooding in the VLAN. Switches running firmwares from 05.17 to 05.61.

Now - at the time the customer opens office, I get a CPU load of ~99%. I have noticed that disabling one of the customer VLANs results that the switch behaves normally. Of course, during high CPU load the switch is sluggish, and respond with one line every 10 seconds or so.

What could trigger this behaviour? Counters state that there are no bc-storm;

v02ravs1# clear statistics 1
(...)
v02ravs1# sho int 1

Status and Counters - Port Counters for port 1

Name :

Link Status : Up

Bytes Rx : 21,512,874 Bytes Tx : 52,093,343
Unicast Rx : 51,402 Unicast Tx : 59,421
Bcast/Mcast Rx : 13 Bcast/Mcast Tx : 355

FCS Rx : 0 Drops Tx : 0
Alignment Rx : 0 Collisions Tx : 0
Runts Rx : 0 Late Colln Tx : 0
Giants Rx : 0 Excessive Colln : 0
Total Rx Errors : 0 Deferred Tx : 0

v02ravs1#

[OLARU Dan]

check your topology for loops.
if you're using STP, put all clients (PCs servers and printers) as EDGE in your STP options.

Cheers,
Dan

[Paulen]

This is of course the first thing I check. Now, since these switches are connected to typically 5304xl, I'd at least have some broadcast. And since the switch is connected to a router, it couldn't be loop there.

This happens only with 2512/2524. Can't imagine they run out of MAC address space.

I have logged this with HP - will update this if I get a plausible explanation.

[Mohieddin Kharnoub]

Hi

Adding to OLARU, the loop is in one of the customers Vlan tat you disabled.

Again check how can this Vlan form a loop in your network.

Enable spanning tree anyway on the switches to protect your network.

Enabling Spanning-Tree in a big topology should be planned and properly configured.

If you can explain about the topology you have, and switch models, then you can find the help you need to configure this spanning tree here.

Good Luck !!!

[Matt Hobbs]

Are you running PCM+ by any chance? With PCM+ 2.2 it automatically enables traffic monitoring - on the 2500 series they don't quite have the power to enable 'sampling' on all ports and you may see higher than usual CPU utilisation (although I wouldn't expect 99%).

2500's will also go to 99% for about 20 minutes or so when you first power them on while they generate SSH keys. This is a low priority task though so shouldn't case any performance issues.

In this instance, since you say disabling one of the customer VLANs seems to fix this, I would be trying to see if there is anything unusual happening on that VLAN.

[Marcus J]

Hi,

I had the exact same problem a few weeks ago. used about 15vlans on the 2524 switches, cpu was constantly around 90-99%. and admin interface very slow.
Contacted hp about this but they did not find any solution since the 2524 does not have the "show tech all". We run PCM 2.2+ if that is the problem??
Anyway i had to replace the 2524 with 2650 to get rid of the problem. if it would have been a loop of some kind the same problem should show on the 2650, but they worked fine.

[Paulen]

Hi,

I recon I have to do an exchange too, but I'd like to know what this comes from. Guess I'll have to pull out my OmniPeek and sniff some traffic.

[Paulen]

Came to think of one thing - it seems to be related to torrent traffic - anyone else experienced this? Will try this more thoroughly under this weekend.

[Hp 5300 Went Down]

We also experience this problem, but the cpu remains at 99% also with no cable connected

[Marcus J]

Any updates on this?, im still having issues with this