2011-07-28 06:09 AM
We have come across an issue when connecting to our Netapp Filer from any HP servers configured with HP network interface teaming. We have found the performance to be very poor and the issue is very intermittent. We struggled diagnosing this for some tume, we then found that disabling the HP teaming resolves this issue. We have since discovered that the HP teaming will work fine if you set the 'options ip.fastpath.enable' on the filer to off.
Has anyone else come accross this issue, are we likely to face any other issues from disabling IP fastpath?
2011-07-28 07:35 AM
ip.fastpath.enable ensures that the filer responds back on the same interface that an incoming request comes in on. That means disabling fastpath causes the filer to rely solely on the routing tables;make sure all of your routes are correctly configured on the filer and you should have no issue. I have not had that issue with any of the servers we have connected. Almost all of our servers use HP NIC teaming and none of them have shown any adverse issues. We have fastpath enabled on all eight of our filers.
I remember an issue at a previous employeer about issues with a Cisco Nexus 10GbE network and fastpath. I know we disabled it until patches were applied, but I don't remember if they were to the switch or the clients.
2011-07-28 07:39 AM
Thanks for the info Scott, these particular servers are HP DL380 G4s (quite old now I know). I have tried all the latest support packs and firmware updates but still have performance issues with the Netapp, the web interface is even too slow to operate. No problem connecting to other devices on the network. As soon as I disable fastpath it works perfectly.
All seems to be working fine with fastpath disabled so far.
I actually found another article on the web earlier with the exact same issue, see link below. It seems to reference a Netapp KB but this does not really give much info.
2011-07-28 07:55 AM
We experienced the same problem when opening Snap Manager on our HP server running windows 2008. Snap manager was taking 10-15 minutes to enumerate the disks and the problem was with the HP teaming. We change the config of our Cisco switch ports and the HP teaming to use 802.3ad Dynamic with Fault Tolerance and not Automatic. Problem solved.
You may notice a differnace just by configuring the teaming to NFT...?