High CPU consumption problem in Core Switch

[GRAMTL]

my switch core has high CPU usage every 3 minutes, switch logs attached.

when there is high CPU power connection is lost

P.S. This thread has been moved from Comware-Based to ProCurve / ProVision-Based. -HP Forum Moderator

[Pete W]

Do the outages/CPU spikes occur at the same time as the log entries appear such as :

  00828 lldp: PVID mismatch on port C16(VID 1) with peer device port 48(VID unknown)

Regards,

Pete

[GRAMTL]

Thank you,

Fix the problem indicated. But the problem persists, the switch logs attached

[Pete W]

Can't see anything glaringly wrong.

Suggestions :

• When you say "every 3 minutes", this sounds like it may be a L2 loop in the network being caught by BPDU Guard or Loop Guard (not necessarily on this switch) and the interface is being temporarily shut down, and then re-enabled every 3 minutes. I would get onto the console and use "show interface port-utilization" as a base-line, and also when the switch is "busy". See whether there are any interfaces showing substantially higher TX or RX utilization.

• Possibly look at the release notes for K.15.12.0012 - I've seen some very strange issues with 10GbE module crashes in earlier versions of K.15.12.XXXX.

Regards,

Pete

[Pete W]

BTW, this post is in the wrong category. It should be in  "ProCurve / ProVision-Based" and not Comware-Based.

Pete

[Apachez-]

If it were a comware device then I found this in the comware troubleshooting guide which might be helpful:

Troubleshooting other problems

High CPU Usage

To resolve the problem:

1. Locate the tasks that are using CPU resources.
Use the display cpu-usage task [ slot slot id ] command in hidden view to display CPU usage information on cards. Locate the tasks that are using CPU resources.

2. Analyze the tasks that are using CPU resources.
The following are common tasks that use CPU resources:

* bRX1 and bRX2
If these tasks are using CPU resources, the switch might be receiving large amounts of packets that need to be forwarded to the CPU. Do the following:

- Use the display interface command to see whether interfaces have received large amounts of broadcast packets.  
- Check whether a loop exists on the network. If a loop exists, examine the connections between the devices.

* MSTP
Use the display stp bpdu-statistics interface command to view the statistics for received and sent BPDU packets on ports. Examine the counts of TCN/TC packets. If the switch receives a large quantity of TCN/TC packets generated by STP, the switch has to delete MAC address entries frequently. This results in high CPU usage.  

* AGNT
Examine the network management software. Frequent MIB information browsing might result in high CPU usage.

* ROUT
Use the display ip routing-table command to examine route changes. Route flapping might cause high CPU usage.

* L2X0 or L2X1 MAC address synchronization
Examine whether the network topology is stable. If the network topology is not stable, the device has to use a lot of CPU resources to update its MAC address entries frequently.

3. If the problem persists, contact HP Support.

_hidecmd
Enters hidden view.

display cpu-usage task [ slot slot id ]
Displays CPU usage information. This command is available in hidden view.

display interface
Displays Ethernet interface information.

display ip routing-table
Displays brief information about active routes in the routing table.

display stp bpdu-statistics interface
Displays the BPDU statistics for a port.

[GRAMTL · Accepted solution]

Thank you all, the problem was in the snmp monitoring which caused high CPU consumption