Teaming Recomendations/Issues

[fishmn]

Greetings,
We've implemented TLB with Round Robin load balancing as our teaming standard. We felt that we did adequate testing on our redundant network with file transfer and ICMP testing to validate the teaming config.

We are still early on into our production implementation and are running into certain issues that appear to be teaming related. We've had netlogon, kerbero's and userenv errors where the computer account appears to get corrupted and can't talk on the domain until it is rebooted. We had an outage of our production MS LCS server where the MS tech looked at it and the first thing he turned off was teaming and SP2 scalable networking components (which one fixed it I don't know)

I'm speculating that since packet order is not guaranteed with round robin LB may be contributing to if not causing these issues. As is typcial with most of these issues, you can't really load test it short of production use.

Anyway, I'm looking for any experiences/recommendations with TLB/RR. I'm considering recommending that we go with fault tolerant to overcome this but it would be nice to have somethin concrete to show that TLB/RR is bad idea for many apps.

Thanks for any replies.
Ron

[Phil.Howell]

What model server is this?
and what network adapters?
Phil

[fishmn]

Various models all with multifunction NIC's standard on the system board. DL385/585G2, BL 465/685G1. We have included the latest multifunction NIC drivers (3.7.19)

[Phil.Howell]

see this thread about problems when using the NC373i adapters
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1158952
We use a fault tolerant setup with each network port connected to a different switch and the failover works well.
Load balancing should only be needed if you are pushing the limits on a single port, if you do use load balancing, then the doco suggests that you should only use round-robin for webservers and udp traffic, otherwise set it as auto.
Phil