HTTP redirect on Local Director

[p.connah]

I'm using a Local Director 416 v4.2.4 and have a problem using http redirect. We were using generic sticky sessions but soon want requests to come via 1 proxied address - SSL sticky doesn't work too well because IE has a habit of creating a new SSLID all the time.

I have a virtual server 10.1.1.2 and real servers 10.1.1.21,22,23 (web01, web02, web01). As I understood it (!) the local director would see a request come into www and redirect the client browser to web01, web02, web03. I have registered the web01 etc with DNS although they're all on the same address - this is fine under generic sticky using bind, of course.

I have;

virtual 10.1.1.2:443:0:tcp is

url web01 https://web01.site.co.uk/%p

url web02 https://web02.site.co.uk/%p

url web03 https://web03.sire.co.uk/%p

direct-ip 10.1.1.21:443:0:tcp is

direct-ip 10.1.1.22:443:0:tcp is

direct-ip 10.1.1.23:443:0:tcp is

link web01 10.1.1.21:443:0:tcp

link web01 10.1.1.22:443:0:tcp

link web01 10.1.1.23:443:0:tcp

backup 10.1.1.21:443:0:tcp 10.1.1.2:443:80:0:tcp

backup 10.1.1.22:443:0:tcp 10.1.1.2:443:80:0:tcp

backup 10.1.1.23:443:0:tcp 10.1.1.2:443:80:0:tcp

When I go to www.site.co.uk, would I expect to see my browser redirect to web01.site.co.uk. Also the documentation on the release notes is a bit vague - it gives the impression that you need to register the INTERNAL addresses on DNS, surely I must have misunderstood this - if the browser does redirect to web01 and I've got that registered as 10.1.1.21 instead of the external address surely the client will instantly hit 'page not found'.

Any information would be appreciated as I can't find much on this. Also any debugging tips would be a help.

Thanks, Paul

[cschneid]

Perhaps the following link would help:

http://www.cisco.com/warp/public/117/local_director/http2https.html

Connections to the main virtual (should be port 80) will get redirected to one of

the three names, web01, web02, or web03. Each one of these names should be registered in DNS as a unique IP address. The reason for this is to ensure that a user is stuck a particular server for the duration of their session.

[p.connah]

Hmm.. This seems a bit stupid. It means all of a sudden I have to have an external IP address for all my Web servers - what happened to the concept of virtual servers, NAT etc? Why can't the load balancer still handle using a virtual address seeing as it knows that Web01 = 10.1.1.21, Web02 = 10.1.1.22 etc.

This seems so wrong it can't be right - can it?

Thanks for the reply btw!