NLB and TS Gateway - failover works, loadbalancing doesnt:

  • Thread starter Thread starter KG
  • Start date Start date
K

KG

Guest
Hi there,



I've got NLB working as a front end to a ts gateway farm. It does failover
magnificently. Connections get reconnected to another ts gateway server
when the one with the highest priority drops off. I would also like the TS
gateways to be load balanced, however, I am not experiencing this in my
environment.



Instead, all connections are sent to the server with the higher priority
(the first one added to the cluster), unless that server is unavailable.
Then they get sent to the second.



I have noticed that all my connections to the gateway "look" like they are
coming from the firewall IP address because they are all coming in from
remote locations. Does that have anything to do with the fact the failover
works but load balancing is not?



All servers run Windows Server 2008.



IP addresses of the gateways and cluster IP are:

x.x.x.22 = TS Gateway 1

x.x.x.34 = TS Gateway 2

x.x.x.88 = Cluster IP



Some NLB specifics:

My rule is for load balancing port 443, both TCP and UDP

Filter mode = Multiples host

Affinity = None (although i have tried Single with the same results)

Cluster operation mode: multicast (although I have tried Unicast too and it
works the same way)



Is this expected behavior for using NLB with TS Gateways? Or is this
solution supposed to produce load balancing and just not doing it in this
case?



Thanks,



KLG
 
Re: NLB and TS Gateway - failover works, loadbalancing doesnt:

Sorry, nevermind. I answered my own question.

NLB of course does IP Affinity. I had a affinity set to "none". Once I set
it to Single, load balancing commensed.

The addresses appearing to be form teh firewall does play a part in this as
they all look to NLB like they are coming from the same IP address. So with
affinity turned on, they go to the same TSGateway. I resolved this with
DNS.

I did have issues with 64bit HyperV NICs suddenly not accepting the cluster
IP addition to the TCP/IP stack. Removing and re-adding the vitual NICs did
not help. I have to switch to another pair of NICs that had not yet been
used for NLB, and then rebuild. Anyone else have that issue?

Cheers,

KG

"KG" <donotreply@msn.com> wrote in message
news:%231$GkHmyIHA.1240@TK2MSFTNGP02.phx.gbl...
> Hi there,
>
>
>
> I've got NLB working as a front end to a ts gateway farm. It does
> failover magnificently. Connections get reconnected to another ts gateway
> server when the one with the highest priority drops off. I would also
> like the TS gateways to be load balanced, however, I am not experiencing
> this in my environment.
>
>
>
> Instead, all connections are sent to the server with the higher priority
> (the first one added to the cluster), unless that server is unavailable.
> Then they get sent to the second.
>
>
>
> I have noticed that all my connections to the gateway "look" like they are
> coming from the firewall IP address because they are all coming in from
> remote locations. Does that have anything to do with the fact the
> failover works but load balancing is not?
>
>
>
> All servers run Windows Server 2008.
>
>
>
> IP addresses of the gateways and cluster IP are:
>
> x.x.x.22 = TS Gateway 1
>
> x.x.x.34 = TS Gateway 2
>
> x.x.x.88 = Cluster IP
>
>
>
> Some NLB specifics:
>
> My rule is for load balancing port 443, both TCP and UDP
>
> Filter mode = Multiples host
>
> Affinity = None (although i have tried Single with the same results)
>
> Cluster operation mode: multicast (although I have tried Unicast too and
> it works the same way)
>
>
>
> Is this expected behavior for using NLB with TS Gateways? Or is this
> solution supposed to produce load balancing and just not doing it in this
> case?
>
>
>
> Thanks,
>
>
>
> KLG
>
>
>
>
>
 
Back
Top