Arp Problem with Linux host

[Axel S.]

Hello guys,

i have a problem with our cisco 3850 when connecting a linux host. The host gets its IP via DHCP as supposed and works just normal as long as traffic is caused on the network interface. After 10 minutes waiting and no traffic is caused on the interface, the host will not longer be pingable. Even tho the host still shows a connection and still has its IP. We found out that a clear ip arp on our core Switch (4500e) helps until the host runs into another "10-minute-pause". 

So my question is now: is this problem known and is there any real solution beside clear ip arp for it?

Our 3850 runs      03.06.06E         cat3k_caa-universalk9

and the 4500e  runs     cat4500e-UNIVERSALK9-M, Version 03.08.02.E

i hope u can help me and thank u in advance

Regards Axel

[brselzer · Accepted solution]

Hello Axel,

It sounds like you might be hitting this issue on the 4500:

https://bst.cloudapps.cisco.com/bugsearch/bug/CSCvb78700

If the 4500 doesn't learn the mac address, it won't flood out all ports. If the server has been inactive for 10 minutes this would cause us to age out the mac address. You can test this by clearing the mac address on the 4500 and see if that breaks it until you learn the mac again:

clear mac address dynamic address [mac address of server]

If you want, I can confirm this is the issue you are hitting if you can post the following two outputs:

show plat hard floodset vlan [vlan of server]

show plat soft floodset vlan [vlan of server]

Thanks!

[Axel S.]

Hi Bradley,

since we restarted the system over the last night, the problem seems to be solved. The bug Report u mentioned seems to be the correct one. Its describtion sounds like what we had. We will force our NOS guy to fasten up, so we can update the 4500 to a release where the issue is fixed. If the problem will occure in the future again, i'll write here again.

Thank you for your fast help.

Axel