Network and Storage Protocols

Odd networking behavior (OnTap 7.3.6) - arp ok, ping not okay

JIM_SURLOW
3,042 Views

I have a situation where I have had a functional network for the filer - single vif, multiple vlans.  Now, hosts cannot mount off one of the subnet ranges off one of the vlans.

I thought that this problem was occurring only with some initial testing that I did with ipspace and vfilers.  But, now the problem has occurred in our vFiler0 environment and default ipspace.

Since the mounts were active, the filer has the MAC information in the arp table.  And since the filer is homed on the vlan with the NFS clients, there isn't any Layer 3 routing going on (I need to find out if for this latest case, if the NFS clients are on the same switch). 

I can ping the subnet gateway, but I cannot ping the NFS clients on the subnet (and of course, all of the NFS mount points are hung). 

Netstat shows me "syn_rcvd" (and not "established") for those NFS clients.

When this problem occured during my ipspace and vfiler testing, a packet trace showed that pings came in and responses went out (client didn't receive them) and outbound pings from the filer went out with no response.  I'm going to guess that I see similar behavior.  (In this initial case, I know that the filer is connected to one switch and the NFS clients were connected to a 2nd switch with an interconnect between the two Cisco switches). 

Running OnTap 7.3.6 - didn't see any networking bugs similar to this in the 7.3.7RC1 bug fix list. 

Anyone have any ideas or thoughts?

3 REPLIES 3

JIM_SURLOW
3,042 Views

And DNS is not a factor here as the connectivity between server & client is all IP number-based, no names.

aborzenkov
3,042 Views

Duplicated IP sounds like quite possible cause.

JIM_SURLOW
3,042 Views

Leading theory at this point is that there is a bad ASIC on the switch.  Data is headed outbound from the filer(s) to the switch.

Public