I'm facing a problem with the ExtremeXOS switch and need your help/advice!
The situation is, my host just made a change in the MAC address for its IP (i.e. IP migration), it immediately sent out GARP packet to notify the switch. Unfortunately, for a certain reason, the switch looks like got a delay of ~650ms before updating the new MAC address for this host IP in its ARP cache. The evidence is, I saw all the ingress traffics arrived in the switch from outside were being routed the old MAC address incorrectly during this period!!! This caused a lot of retransmissions on the network, and in the worst case the connection/association towards the host would be aborted.
How to overcome this?
Can we configure the switch so that it will update the ARP cache right when receiving GARP packets?
Here are the "show iparp" printouts from the switch:
Dynamic Entries : 247 Static Entries : 0
Pending Entries : 0
In Request : 9767973 In Response : 5676233
Out Request : 5676564 Out Response : 4413793
Failed Requests : 44
Proxy Answered : 1081698
Rx Error : 0 Dup IP Addr : 0.0.0.0
Rejected Count : 296491 Rejected IP : 192.168.169.56
Rejected Port : 1 Rejected I/F : BSC_Inner
Max ARP entries : 8192 Max ARP pending entries : 256
ARP address check: Enabled ARP refresh : Enabled
Timeout : 20 minutes ARP Sender-Mac Learning : Disabled
Locktime : 1000 milliseconds
Retransmit Time : 1000 milliseconds
Reachable Time : 900000 milliseconds (Auto)
I'm not sure what the real meaning of the "Locktime" parameter is designed by the ExtremeXOS (version 126.96.36.199)? Could it be the root cause? Note that, the MAC address' change only happened once in this case (i.e. not repeatedly/frequently).
Thanks for your reply!
But as my understanding, the locktime is usually to prevent situations where a new change for the same ARP entry is too fast since the last change (i.e. the interval between the two changes is not allowed =< 1000ms with this definition)?
Is there a difference in the way that the locktime is defined by ExtremeXOS?