Does the NLB Service 'Know' if Something is Listening?

[Guest Charles Law]

When the NLB service forwards to the nodes in an NLB cluster, does it know

whether something is listening on a given port?

 

Suppose I have a 2-node NLB cluster. I run the same program on both nodes,

listening on the same port. I expect that NLB will share incoming

sessions/messages between the nodes, depending on how I have configured it.

 

When a node goes down, NLB should just send messages to the remaining node.

 

But what happens if both nodes are running, and I stop my program running on

one of the nodes? Is NLB clever enough to see that there is nothing

listening to the port on one of the nodes, and therefore just forward to the

other node, or does it continue to share regardless?

 

Thanks for any insights.

 

Charles

[Guest Russ Kaufmann [MSFT]]

Re: Does the NLB Service 'Know' if Something is Listening?

 

"Charles Law" <blank@nowhere.com> wrote in message

news:OANmfzEtIHA.420@TK2MSFTNGP02.phx.gbl...

> When the NLB service forwards to the nodes in an NLB cluster, does it know

> whether something is listening on a given port?

>

> Suppose I have a 2-node NLB cluster. I run the same program on both nodes,

> listening on the same port. I expect that NLB will share incoming

> sessions/messages between the nodes, depending on how I have configured

> it.

>

> When a node goes down, NLB should just send messages to the remaining

> node.

>

> But what happens if both nodes are running, and I stop my program running

> on one of the nodes? Is NLB clever enough to see that there is nothing

> listening to the port on one of the nodes, and therefore just forward to

> the other node, or does it continue to share regardless?

 

NLB does not do application level health checking. What you can do, however,

is implement a monitoring solution and then the monitoring solution can run

a command or script to restart the application if it is found in a failed

state.

 

--

Russ Kaufmann, MCSE, MCT, MCITP, MCTS, and other stuff

Premier Field Engineer - Exchange

Microsoft Services

[Guest Charles Law]

Re: Does the NLB Service 'Know' if Something is Listening?

 

Hi Russ

 

Thanks for the reply. It's not so much health-checking that I am after, but

I want to prevent loss-of-service.

 

When I come to update my program, I would like to stop it on one node, and

start the new version. Whilst this is going on, I want the program on the

other node to take over completely, so that the service is maintained. When

the new version is running I will do the same on the other node, and my

program will again run on both nodes.

 

Charles

 

 

"Russ Kaufmann [MSFT]" <russ.kaufmann@discussions.microsoft.com> wrote in

message news:1A1785BD-8A9F-4651-A52C-7C746B4497F5@microsoft.com...

> "Charles Law" <blank@nowhere.com> wrote in message

> news:OANmfzEtIHA.420@TK2MSFTNGP02.phx.gbl...

>> When the NLB service forwards to the nodes in an NLB cluster, does it

>> know whether something is listening on a given port?

>>

>> Suppose I have a 2-node NLB cluster. I run the same program on both

>> nodes, listening on the same port. I expect that NLB will share incoming

>> sessions/messages between the nodes, depending on how I have configured

>> it.

>>

>> When a node goes down, NLB should just send messages to the remaining

>> node.

>>

>> But what happens if both nodes are running, and I stop my program running

>> on one of the nodes? Is NLB clever enough to see that there is nothing

>> listening to the port on one of the nodes, and therefore just forward to

>> the other node, or does it continue to share regardless?

>

> NLB does not do application level health checking. What you can do,

> however, is implement a monitoring solution and then the monitoring

> solution can run a command or script to restart the application if it is

> found in a failed state.

>

> --

> Russ Kaufmann, MCSE, MCT, MCITP, MCTS, and other stuff

> Premier Field Engineer - Exchange

> Microsoft Services

[Guest Russ Kaufmann [MSFT]]

Re: Does the NLB Service 'Know' if Something is Listening?

 

"Charles Law" <blank@nowhere.com> wrote in message

news:uPJyrDNtIHA.1436@TK2MSFTNGP05.phx.gbl...

> Hi Russ

>

> Thanks for the reply. It's not so much health-checking that I am after,

> but I want to prevent loss-of-service.

 

That is a matter of definition. What I mean by that is that we are basically

saying the same thing. By health checking, I mean that it is verifying the

service is running properly and is thus, healthy. :)

 

Sorry if I was not clear.

> When I come to update my program, I would like to stop it on one node, and

> start the new version. Whilst this is going on, I want the program on the

> other node to take over completely, so that the service is maintained.

> When the new version is running I will do the same on the other node, and

> my program will again run on both nodes.

 

Ah, I see. So, what you can do is drain stop one node, perform maintenance,

bring it online, then drain stop the other node and repeat.

 

Hopefully this link of the commands will help:

http://technet2.microsoft.com/windowsserver/en/library/4d7051b5-0985-42d3-becc-4d8f1ca0cedd1033.mspx?mfr=true

 

--

Russ Kaufmann, MCSE, MCT, MCITP, MCTS, and other stuff

Premier Field Engineer - Exchange

Microsoft Services

[Guest Charles Law]

Re: Does the NLB Service 'Know' if Something is Listening?

 

Thanks for the link. When I drain stop a node and then bring it back

on-line, is there a convergence period when both nodes appear to be

temporarily off-line? I have noticed this when changing the config of either

node, both nodes adopt a status of converging... which suggests to me that

they might not be processing during that time, albeit briefly.

 

Charles

 

 

"Russ Kaufmann [MSFT]" <russ.kaufmann@discussions.microsoft.com> wrote in

message news:B54FD689-A076-4E55-AACF-0C3F944C9378@microsoft.com...

> "Charles Law" <blank@nowhere.com> wrote in message

> news:uPJyrDNtIHA.1436@TK2MSFTNGP05.phx.gbl...

>> Hi Russ

>>

>> Thanks for the reply. It's not so much health-checking that I am after,

>> but I want to prevent loss-of-service.

>

> That is a matter of definition. What I mean by that is that we are

> basically saying the same thing. By health checking, I mean that it is

> verifying the service is running properly and is thus, healthy. :)

>

> Sorry if I was not clear.

>

>> When I come to update my program, I would like to stop it on one node,

>> and start the new version. Whilst this is going on, I want the program on

>> the other node to take over completely, so that the service is

>> maintained. When the new version is running I will do the same on the

>> other node, and my program will again run on both nodes.

>

> Ah, I see. So, what you can do is drain stop one node, perform

> maintenance, bring it online, then drain stop the other node and repeat.

>

> Hopefully this link of the commands will help:

> http://technet2.microsoft.com/windowsserver/en/library/4d7051b5-0985-42d3-becc-4d8f1ca0cedd1033.mspx?mfr=true

>

> --

> Russ Kaufmann, MCSE, MCT, MCITP, MCTS, and other stuff

> Premier Field Engineer - Exchange

> Microsoft Services

[Guest Russ Kaufmann [MSFT]]

Re: Does the NLB Service 'Know' if Something is Listening?

 

"Charles Law" <blank@nowhere.com> wrote in message

news:%23ZsmG5QtIHA.420@TK2MSFTNGP02.phx.gbl...

> Thanks for the link. When I drain stop a node and then bring it back

> on-line, is there a convergence period when both nodes appear to be

> temporarily off-line? I have noticed this when changing the config of

> either node, both nodes adopt a status of converging... which suggests to

> me that they might not be processing during that time, albeit briefly.

 

Yep, they have to re-converge, but once that happens, then traffic should

start moving to the node.

 

 

--

Russ Kaufmann, MCSE, MCT, MCITP, MCTS, and other stuff

Premier Field Engineer - Exchange

Microsoft Services