Routing

Hi.

I have a problem where two EX9200 switches  configured in MC-LAG do not seem tp replicate their ARPs between each other.

The situatio is:

Someone deploys a VM and is assigned an IP

They then destory that VM and deploy and new one and they get the same IP but a differnet MAC.

The new IP/MAP (ARP) is learnt on one switch, but it does not get synchroinzed to the other switch.

Here is the output:

So here on switch2, you can see the ARP changes automatically 

me@switch2> show arp hostname 192.1.1.1 no-resolve    
May 27 01:05:16
MAC Address       Address         Interface     Flags
04:01:1a:53:31:01 192.1.1.1  ae62.0               none

{master}
me@switch2> show arp hostname 192.1.1.1 no-resolve    
May 27 01:05:28
MAC Address       Address         Interface     Flags
04:01:1a:53:36:01 192.1.1.1  ae62.0               none

But on switch1 it takes about 6 minuyes to change... see the timestamps on each show command

me@switch1> show arp hostname 192.1.1.1 no-resolve    
May 27 01:02:47
MAC Address       Address         Interface     Flags
04:01:1a:53:31:01 192.1.1.1  ae62.0               none

{master}
me@switch1> show arp hostname 192.1.1.1 no-resolve    
May 27 01:02:51
MAC Address       Address         Interface     Flags
04:01:1a:53:31:01 192.1.1.1  ae62.0               none

{master}
me@switch1> show arp hostname 192.1.1.1 no-resolve    
May 27 01:03:00
MAC Address       Address         Interface     Flags
04:01:1a:53:31:01 192.1.1.1  ae62.0               none

{master}
me@switch1> show arp hostname 192.1.1.1 no-resolve    
May 27 01:03:29
MAC Address       Address         Interface     Flags
04:01:1a:53:31:01 192.1.1.1  ae62.0               none

{master}
me@switch1> show arp hostname 192.1.1.1 no-resolve    
May 27 01:04:58
MAC Address       Address         Interface     Flags
04:01:1a:53:31:01 192.1.1.1  ae62.0               none

me@switch1> show arp hostname 192.1.1.1 no-resolve     
May 27 01:08:16
MAC Address       Address         Interface     Flags
04:01:1a:53:36:01 192.1.1.1  ae62.0               none

So it finally changes ater about 6 minutes.

Any idea why this is happening.  ICCP should be replicating this.

Running Junos 12.3R5.7, EX9208.

I did notice this in the release notes... It could be related:

On EX9200 switches, dynamic ARP resolution is not supported over inter chassis control
links(ICLs). As aworkaround, configure static ARP on both ends of the ICL.[PR/850741:
This is a known software limitation.]

sounds like you may be running into one (or both) of these PRs:

PR961479:

******

On EX4300/EX9200 Series switches with Dynamic Host Configuration Protocol (DHCP) relay configured, permanent Address Resolution Protocol (ARP) entries for relay clients are installed. When the client is reachable via a different preferred path (due to STP topology changes / MC-LAG changes etc) the forwarding state is not refreshed. This might cause the traffic to be blackholed until the relay binding is cleared.

Workaround:

Add the following configuration to the forwarding-options stanza to suppress installation of destination routes (DHCPv4 only):
root@switch> set forwarding-options dhcp-relay route-suppression destination

PR901361

On EX9200 Series switches with Multichassis Link Aggregation Group (MC-LAG) interface configured, if one link of MC-LAG flaps, the layer 3 traffic over MC-LAG interface might be dropped because ARP entries are not pointing to the correct next hop.

Workaround:

As a workaround, please enable "arp-l2-validate" knob by the following command:
root@switch# set interfaces irb arp-l2-validate

Note, this is a hidden command.

please report back if you resolve the issue!

Thanks,

will

Thanks for info.

Not sure if they are related thought.  We are not running DHCP relay on the switches so PR961479 is probably not relevant.

Also the MC-LAG interfaces do not flap when this occurs so not sure is PR901361 is the issue also.