I have a problem with an ESX3.5 host being able to ping an NFS storage device which holds the datastores. I have two NFS NetApp storage devices S100 and S101, the ESX host can vmkping S101 ok but cannot ping S100. I have done some quite detailed investigation into this and it seems that the ESX host is not updating the vmkernel arpLookup table. using tcpdump i can see the arp request and the arp reply however the mac address is not stored in the arpLookup by the ESX host.
Also sometimes the vmkping gives the following error
sendto() failed (Input/output error)
has anyone seen this kind of behaviour or could anyone give any advice, on the ESX side i have a vSwitch configured containing a vmkernel and service console both using multiple physical interfaces with ip hash selected. These are connected to a cisco switch with a port channel configured. I have discussed the issue with the network team who cannot find a problem and since the arp reply gets back to the ESX host then i tend to agree with them.