I'm having trouble figuring out the best way to utilize redundant connections on several of our Windows based servers. We're running a large cluster of powerEdge 750s which come with dual gigabit NICs. Some of the servers run Linux. For those servers its a cake walk. We configure LACP between the servers and our access layer switches (ie utilizing both physical interfaces) so we not only gain bandwidth but can utilize the failover capabilities of the protocol, and in addition, we get a virtual IP address. I wish this was easy to do on Windows. Yes I know Windows (via NIC drivers such as Intel PRO) supports LACP but its total crap and we actually dont seem to have that capability on the PowerEdge 750s.

I have done it on other PowerEdge servers and it always glitches or performance stinks. So im trying to think of other ways to simulate this on these existing servers. I wish I could just throw Windows out all together but its a slow process to get off old hardware. Also, virtulization is not a possibility with these servers. I wish they were 64 bit and could run vSphere but we're stuck doing things the old fashioned way for the time being until we can push out this old hardware.

Anyways, I thought about maybe assigning two IPs, (one to each interface) and then doing some sort of DNS alias or round robin technique, that way we can access the server via its hostname and it will use one of the available IPs. Whats the best way to go about configuring this? CNAME on DNS? Round Robin? Some other type of load balancer technology? Trying the IntelPRO LACP drivers again (if they even exist for the poweredge 750s) or something else im not thinking of. Please help. Thanks

Oh I guess I should add, these Microsoft boxes are still running Server 2003. Yes hush, we know.

  • Don't take this the wrong way but if you don't have a specific need or use case for binding/linking/aggregating the NIC's on the Windows servers then it seems counter-productive to me to do so. Just because the servers have more than one NIC or NIC port doesn't mean you have to put them both/all to work. Again, no offense but work done without having a specific need, reason or problem that needs solving is just so much "busy work" IMO. – joeqwerty Aug 6 '13 at 23:41
  • 1
    Wow, could you please add some paragraphs to your text so that it is actually readable? – Frederik Nielsen Aug 6 '13 at 23:41
  • 2
    Just because you can put 5 tags on a question doesn't mean you should. – longneck Aug 6 '13 at 23:42
  • Are you really solving/accomplishing anything by pursuing this course of action? I don't do anything "just because". I make sure I have a specific reason for doing it and a specific need to be filled or problem to be solved. The servers I manage all have redundant NIC's and they're all currently unused and disconnected. – joeqwerty Aug 6 '13 at 23:42
  • Yes its called redundancy for mission critical servers. I appreciate all of the useful feedback too... – Atomiklan Aug 6 '13 at 23:45

If these are truly mission critical servers which need the redundancy get a supported combination of NIC, driver, OS and switch so LACP or some other form of failover works properly. Not sure what you will have to change since your diagnosis of trying it amounts to "its total crap".

Otherwise leave one disconnected. Or run the extra cable but leave the nic disconnected so you can easily switch it without touching the wiring.

For what its worth I have several PowerEdge servers ruunning both Windows 2003 and 2008 and using both broadcom and intel cards and have had them running in teams without any issues, though I did have to try several different teaming modes to get it running smoothly.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.