Linux hosts randomly unreachable until arp cache is manually cleared.

[KosstDukat]

Hello all,

Running into an issue that I've not been able to resolve yet.  In my environment, I have two Linux devices which, either after a reboot or after a few hours of not accessing the device directly (the systems run in the background), I am unable to reach them without manually running the 'clear arp-cache' command on the switch.  However, this problem only occurs if the devices are being accessed by a computer that resides on a different VLAN.  If I try to access the devices from a computer on the same VLAN, there is no issue.  This issue does not occur with any of the Windows systems in the environment.  I've tried to set a manual mac binding for these two devices but that didn't fix the issue.  I did find a page with a similar issue that recommended turning gratuitous ARPs on, but the security requirements of my environment do not allow that.

Does anyone have any other ideas?  I'm scratching my head on this one.. 

Thank you!!

[liviu.gheorghe]

Hello @KosstDukat ,

"I am unable to reach them without manually running the 'clear arp-cache' command on the switch" - this switch is the one that handles the inter Vlan routing and all the host on a given vlan have the SVI defined on the switch as their default gateway?

When you observe the described situation again, before clearing the arp cache, can you issue the following commands on the switch:

show ip arp

show mac address-table 

show vlan

[KosstDukat]

Hello @liviu.gheorghe , thank you for your response!

Yes, that's correct.  

Attached is the info you requested (some information redacted for security).

[liviu.gheorghe]

For the two linux hosts, can you provide their IP addresses, last 2 bytes, and the vlan number?

Also, when lack of connectivity happens, can you access the hosts on the console and verify the following:

arp -na

netstat -nr

Can you also provide the output of show ip interface vlan x, the SVI corresponding to the linux host vlan, and show ip route?

[MHM Cisco World]

when issue happened again 

show vlan breif 
share if here 

MHM

[Joseph W. Doherty]

Different VLANs, eh?

Sort of hard to see how the ARP cache, on a L2 switch, would impact two hosts, unless this is also a L3 switch?

However, with certain topologies, you can run into problems with MAC and ARP timers being different, although that often leads to unicast flooding.  Unsure that would cause what you describe, but I wonder.