asked on

NLB IIS 7.5 issues

Hi all, I've got a set of 2 web servers running IIS on server 2008r2 enterprise. These are virtual machines on esx 5.1 hosted by a remote 3rd party datacenter. We are trying to get Windows Network Load Balancing working, as we have a fair amount of traffic coming in, 9000+ connections and 100+ mpbs of bandwidth currently being served by only one of the web servers. When we enable NLB things get screwy, people get logged off or the website(s) become unavailable. This used to work (or so I am told, I inherited this project) Weve been going back and forth with the hosting company for days with no real progress in site and the website is production. Based on my cursory research Windows NLB is or at least can be a real bear. I asked what our alternatives were if we making no real progress without any resolution and he suggested a virtual load balancing appliance by f5 as a much better solution.

I imagine many of you will have a lot more questions about the environment but we would like to keep using win NLB due to it's freeness but should we just stop screwing around with this and get the more resilient virtual appliance? It's a bit spendy...

Andrej Pirman

NLB works on IIS 7.5 just fine, but I guess visitor's sessions are lost, or even worse, you don't have some Reverse Caching Proxy in front, which may lead to misconfigured multipathing issues. By my oppinion, one of the following is the best to use with NLB:
- either cloud-based reverse caching proxy
- or local reverse caching proxy, which does not need to be Microsoft's (you ca use Squid, for example, to load balance incoming requests and make sure the same visitor gets served from the same backend server for the whole session duration)

On the second thought, just check if you have "Sticky Sessions" enabled, called Client Affinity:
enable client affinity in the Add/Edit Port Rules dialog box in Network Load Balancing Manager.
This makes sure the visitor gets served for the duration of session from the same host.

LWDud

ASKER

So I double checked the affinity, it's set properly.

we got through about 2 hrs of operqtion and thengs went haywire. Ive enabled the secondary host and taken primary offline to see if the individual hosts behave this way. It is however likely an NLB issue, as shutting off the secondary host makes the issue go away. Traffic still routing through the cluster ip. Even though NLB is not application aware... I'll report results later tomorrow

Dan McFadden

When you say things go haywire, what happens?

1. Are there any corresponding messages in the Sys/App event log?
2. Are you using a unicast or multicast setup?
3. how many NICs are in the VMs?
4. are there other VMs on the same vSwitch? are they having communication issues as well?

Reference links:
1. http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1006778
1a. https://www.vmware.com/files/pdf/implmenting_ms_network_load_balancing.pdf
2. https://technet.microsoft.com/en-us/library/cc962174.aspx

Dan

LWDud

ASKER

Right so when I say haywire I mean people randomly getting kicked out all the way back to the logon page or a generic .net/iis server application error. We are in multicast mode, the 2 vm's have 2 nics one for heartbeat(private) one for actual traffic(public).

We are in multicast mode, I do have other vm's on the same subnet and other vm's are not having issues with communication but they are also not clustered. I will review your links shortly.

ASKER CERTIFIED SOLUTION

LWDud

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

LWDud

ASKER

There was nothing wrong with NLB. We (the IT team) were pointed in the direction of NLB issues but over time it became clear to us there was something else wrong.