arp request every 90 sec on 6500

[mgoerke]

Hello,

Has anyone an idea, why the Switch 6500 sends every 90 seconds a arp request? The cpu load is nearly 100% while this request occur.

We haven’t any arp timer configured, we use the default timer 4 hr.

- for all arp entry’s the switch will generate a refresh unicast arp request every 90 sec

- so no entry “sh ip arp” is older than one minute

- non responded host will time out after 4 hrs

- hsrp on/off: does not matter

We use four WS-C6509-E, WS-SUP720-3B, ca. 8000 host, ca. 700 Vlans and a lot of vrfs.

What can I do to narrow down this behavior?

Perhaps, a activated service or a overloaded (hw) table?

Thank you!

[mgoerke]

Hello,

we have found the cause,

a flapping server interface every 90 seconds is the culprit:

We use MSTP with different regions,

coupled via (rapid)CST.

(a good mstp-explanation with multiple mstp-regions: http://blog.ine.com/2010/02/22/understanding-mstp/)

The server is connected via a trunk. Hence, every flapping trunk interface cause a IST-Change (MSTI0),

although we haven't any VLANs in MST Instance 0.

This Topology change affect all MSTP-Regions, and, as normal SPT-behaviour, all mac-address-table will clear, unfortunately also all arp-tables.

So, every 90 secound we had thousands of arp-requests.

Thanks for thinking about it,

Morris

[pille1234 · Accepted solution]

Hi,

that's an interesting occurence you describe here. A flapping link on an edge port should not have any impact on spanning tree and no topology change notification should be sent. That is, of course, if you configured the server port as an edge port (spanning-tree portfast trunk).

Regards

Pille

[mgoerke]

Pille,

thank you!

Of course, i tried "spanning-tree portfast" unfortunately without additional "trunk" :-)

Now we will use portfast trunk and bpdu-guard.

Again, thank you.

Morris