HP 2920-48G PoE daisy-chain ARP cache problem

arphp-procurve

We've got two HP2920-48G PoE switches connected together via a simple 1Gb cable (SW2:48 -> SW1:01), and then SW1:A1 fiber-channel to a central switch (SW9:A1).

All switches are untagged in the DEFAULT_VLAN (one /19 broadcast domain).

We have PoE Cisco phones plugged into SW2 and basically PCs/Macs plugged into SW1.

When we do a clear arp in SW2, it seems to be able to ping everything in the network (as expected), but moments later it loses its ability to connect to certain IP addresses within the broadcast domain, including SW1 which is the switch it directly connects to. However, all the phones still work, and anything running ethernet still switches correctly. Its only the access to and from SW2:DEFAULT_VLAN address that seems to go bad.

We are not network engineers, but we know our way around switches a bit. What would be causing bad ARP entries, or why would clear arp fix it momentarily, but only for the moment?

Best Answer

ARP is a layer-3 ("IP") technology. Under normal conditions an ethernet switch should have nothing at all in it's arp cache, because 100% of what it does is layer-2 ("ethernet") switching. The only arp entries would be for management access (eg. the MAC of your machine would be known to it while you have a connection to it.)

If the UI is going away but it's otherwise still passing traffic, that sounds like something is attacking the switch, or otherwise overloading it (eg. thinking it's a router, or other server.)

hp-2810-1# show arp

 IP ARP table

  IP Address      MAC Address       Type    Port
  --------------- ----------------- ------- ----
  192.168.0.1     000423-c6a4ec     Dynamic Trk1 (the router)

vs.

hp-2810-1# show mac-address 

 Status and Counters - Port Address Table

  MAC Address   Located on Port
  ------------- ---------------
  000423-c6a4ec Trk1           
  000423-c6a4ed Trk1           
  .... (many pages, in fact)

Despite the volume of broadcast traffic through the switch, very little (none) is in the management vlan (10 in my case.)

hp-2810-1# show cpu

4 percent busy, from 178 sec ago
1 sec ave: 3 percent busy
5 sec ave: 2 percent busy
1 min ave: 5 percent busy
Related Topic