cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1327
Views
0
Helpful
5
Replies

arp_mgr high cpu

a.altayeb
Level 1
Level 1

Dear

We have Cisco ACE used in our company and yesterday, we had add new servers to loadbalancer (which have teaming configured Active/Active) and after maybe 3 hours of implmentation our network became unstable since we had disconnection and connection again on different VLANs and we noticed that arp_mrg have consumed about 98 % of cpu process.

So, my questions are:

1- Does arp_mgr cause such impact of unstability? and what to do to reduce this process if we see it again?

2- do you know a tool can monitor cisco ACE CPU and error logs and warning ?

I appreciate your fast response

Thank

Abdul Elah

5 Replies 5

jsirstin
Level 1
Level 1

Abdul Elah,

It sounds like the type of teaming you are doing is the problem. The ACE does not play well with Active Active teaming. You should configure your NICs to use an Active/Standby configuration so the ACE will only see a single MAC address for this rserver.

Regards

Jim

So, as result of this teaming and high CPU utilize, am I expecting unstability of ACE connected VLANs ? because it was stable for some time then started to be unstable for all connected VLAN. However, I do believe this might be the issue but I need to find out some information or logs in ACE proves this assumption to support this case.

Also, I am looking to find tool or a way to monitor ACE errors and warning messages and maintain them

Thanks,

Abdul Elah

Abdul Elah,

You will most likly see syslog messages like the one below.

This link discussed the message in a little more detail.

http://www.cisco.com/en/US/docs/interfaces_modules/services_modules/ace/v3.00_A2/configuration/system/message/guide/messags.html#wpxref94708

Error Message   %ACE-4-405001: Received ARP {request | response} collision from
IP_address/mac_address on interface interface_name

Regards

Jim

Gilles Dufour
Cisco Employee
Cisco Employee

This is a known problem with nic teaming in active-active.

We're trying to address this issue

CSCth66757: arp_mgr consume 100% cpu due NIC teaming/arp flood

In the meantime, I would suggest not to use active-active.

Gilles.

Thanks for your response

Review Cisco Networking for a $25 gift card