Balancing intra-server requests

[Simon Wadsworth]

Hi

I have a server enviroment that performs back-end batch processing. A
controlling server posts http requests to a 10-server nlb cluster. Whichever
of the cluster servers handles the message will then start the batch process.
All is well.

These batch processes can also start other batches, or in the event of an
error rerequest the failed job, using the same http-request to the nlb
mechanism. Howerver, when a server within the nlb posts to the nlb IP address
the request always comes back to the server doing the posting.

I can understand this as it would probably see the nlb address as "itself"
and simply "go local".

My question is, is there any way of getting a server with the farm to post
back via the nlb address so that the request can be handled by ANY of the
servers in the farm?

[Richard Ault [MSFT]]

You might try configuring NLB to use multicast and affinity none.
--
Richard Ault
Application Center Product Team

This posting is provided "AS IS" with no warranties, and confers no rights.
Use of any included script samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm

Post by Simon Wadsworth
Hi
I have a server enviroment that performs back-end batch processing. A
controlling server posts http requests to a 10-server nlb cluster. Whichever
of the cluster servers handles the message will then start the batch process.
All is well.
These batch processes can also start other batches, or in the event of an
error rerequest the failed job, using the same http-request to the nlb
mechanism. Howerver, when a server within the nlb posts to the nlb IP address
the request always comes back to the server doing the posting.
I can understand this as it would probably see the nlb address as "itself"
and simply "go local".
My question is, is there any way of getting a server with the farm to post
back via the nlb address so that the request can be handled by ANY of the
servers in the farm?

[Simon Wadsworth]

Richard

Thanks for the suggestion, but we have already tried this. Do you think it
should have fixed the problem or was this just a "try it and see what
happens" suggestion?

Simon

Post by Richard Ault [MSFT]
You might try configuring NLB to use multicast and affinity none.
--
Richard Ault
Application Center Product Team
This posting is provided "AS IS" with no warranties, and confers no rights.
Use of any included script samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm

Post by Simon Wadsworth
Hi
I have a server enviroment that performs back-end batch processing. A
controlling server posts http requests to a 10-server nlb cluster. Whichever
of the cluster servers handles the message will then start the batch process.
All is well.
These batch processes can also start other batches, or in the event of an
error rerequest the failed job, using the same http-request to the nlb
mechanism. Howerver, when a server within the nlb posts to the nlb IP address
the request always comes back to the server doing the posting.
I can understand this as it would probably see the nlb address as "itself"
and simply "go local".
My question is, is there any way of getting a server with the farm to post
back via the nlb address so that the request can be handled by ANY of the
servers in the farm?

[Richard Ault]

Simon,

It was more along the line of a "this is the only way it has a chance of working" suggestion. I've never tried to do this (usually it's desirable to run spawned jobs on the same server so that they can share context). You might try some of the windows newsgroups (I don't see any obvious ones off-hand) and I'll see if some of our NLB folk have any better answers.

Richard Ault
Application Center Product Team

This posting is provided "AS IS" with no warranties, and confers no rights. Use of any included script samples are subject to the terms specified at http://www.microsoft.com/info/cpyright.htm

-----Original Message-----
From: ***@discussions.microsoft.com
Posted At: Tuesday, November 22, 2005 5:11
Posted To: microsoft.public.applicationcenter.nlb
Conversation: Balancing intra-server requests
Subject: Re: Balancing intra-server requests



Richard

Thanks for the suggestion, but we have already tried this. Do you think it should have fixed the problem or was this just a "try it and see what happens" suggestion?

Simon

Post by Richard Ault [MSFT]
You might try configuring NLB to use multicast and affinity none.
--
Richard Ault
Application Center Product Team
This posting is provided "AS IS" with no warranties, and confers no rights.
Use of any included script samples are subject to the terms specified
at http://www.microsoft.com/info/cpyright.htm

Post by Simon Wadsworth
Hi
I have a server enviroment that performs back-end batch processing.
A controlling server posts http requests to a 10-server nlb cluster.
Whichever
of the cluster servers handles the message will then start the batch
process.
All is well.
These batch processes can also start other batches, or in the event
of an error rerequest the failed job, using the same http-request to
the nlb mechanism. Howerver, when a server within the nlb posts to
the nlb IP address the request always comes back to the server doing
the posting.
I can understand this as it would probably see the nlb address as "itself"
and simply "go local".
My question is, is there any way of getting a server with the farm
to post back via the nlb address so that the request can be handled
by ANY of the servers in the farm?