Forum Discussion
HTTP HTTPS Proxy redirect question
Hi to everyone, not really sure of my title neither if this is the right place to post this as I am very new to the F5 community (3 days :) ) and DC and thanks already to all contributors of that site that already helped me out with my LTM implementation. So far got everything I wanted to for my exchange and View environment but I am left with one issue that I can t figure out or found on this site although I am sure this has already been asked.
To make things simple let's say I have 4 websites and 3 webservers. One internal server hosts 2 HTTP sites (A & B) and the two other servers hosts in DMZ a dedicated HTTPS site (C & D). All sites are accessible from inside and outside. At the moment Site A and B are sharing the same public ip. There is a NAT on the firewall to a proxy that has to be phased out that passes the traffic to the internal webserver. Site C and D have there own public ip that also has a NAT but straight to their respective server in DMZ.
Is it possible to put all sites on the same public ip, Nat that ip to an external VIP on the F5 and let the F5 box redirect the traffic to the right webserver ? That would help me clear out the mess I have been left with and free up a lot of public ip that I will soon need for other stuff.
Thanks a lot for any help that you can provide
Seb
Yes you can do this you create two virtuals using the same IP address one listening to port 80 and the other listening to port 443. The port 80 traffic will be sent to the server which will use the HTTP host headers to display the correct site just like now.
The problem comes with the HTTPS site you have two server each hosting the same sites? Is so put them in one pool and add both SSL certs to the Virtual using the link below
http://support.f5.com/kb/en-us/solutions/public/13000/400/sol13452.html?sr=32430737
The problem you will run into is if the client does not support TLS hostname then the LTM will not know which cert to pass back to the client and will pass back the default cert which in your case has a 50% chance of being the correct cert. Now if most of your clients support this you should not have a problem.
- Richard__HarlanHistoric F5 Account
Yes you can do this you create two virtuals using the same IP address one listening to port 80 and the other listening to port 443. The port 80 traffic will be sent to the server which will use the HTTP host headers to display the correct site just like now.
The problem comes with the HTTPS site you have two server each hosting the same sites? Is so put them in one pool and add both SSL certs to the Virtual using the link below
http://support.f5.com/kb/en-us/solutions/public/13000/400/sol13452.html?sr=32430737
The problem you will run into is if the client does not support TLS hostname then the LTM will not know which cert to pass back to the client and will pass back the default cert which in your case has a 50% chance of being the correct cert. Now if most of your clients support this you should not have a problem.
- Seb1180_135325NimbostratusThanks for your answer. The HTTPS sites are not the same ;) Each HTTPS site as his own server.
- Richard__HarlanHistoric F5 AccountIn that case you use the Host base iRule Kevin posted below and it should work just fine.
- Kevin_StewartEmployee
The short answer is yes, but the solutions for HTTP and HTTPS can be slightly different.
Assuming you're going to use standard ports (80 for HTTP and 443 for HTTPS), for HTTP traffic you can do what's commonly called "host-based load balancing", where the load balancer sends traffic to a pool based on the HTTP request URI. Here's an example:
when HTTP_REQUEST { switch [string tolower [HTTP::host]] { "app1.example.com" { pool app1_pool } "app2.example.com" { pool app2_pool } } }
This is probably the easiest method. You could also use an app-based URI if the host name was the same for both applications (ex. www.example.com/app1 or www.example.com/app2). The real beauty of this approach is that you're putting each application (based on IP and/or port) into its own pool. To add additional resources you can simply add new nodes to the pools and actively load balance and monitor the applications.
The above takes care of port 80 HTTP applications using a single IP address. Doing this for HTTPS is little more complicated in that the SSL negotiation is 1) based on the host name, and 2) happens before the HTTP host name can be evaluated. You can use the same exact iRule above, but you also need to configure the SSL properties of the VIP to accommodate multiple host names. Your options are:
-
A "Wildcard" certificate applied to the single client SSL profile applied to the VIP - this is a single certificate with a wildcard host name (ex. (.example.com) that would cover any host name that matches this domain pattern. These certificates tend to be expensive.
-
A "Subject Alternative Name (SAN)" certificate applied to the single client SSL profile applied to the VIP - this is a single certificate with a single subject name, but multiple subject alternative names. Like a wildcard it can support multiple host names, but unlike a wilcard it only supports a specified subset of host names.
-
Server Name Indicator (SNI) - an extension to TLS 1.0 and above that allows the F5 VIP to switch between multiple single cert client SSL profiles based on the server_name attribute sent by the client in the CLIENTHELLO message of the SSL handshake. This requires TLS so some legacy clients can't use this (ie. WinXP and earlier).
All of these options have the effect of allowing you to negotiate SSL with a single VIP, without error, and then use the host-based iRule above to send traffic to the correct application pool.
-
- Seb1180_135325Nimbostratus
Thanks Kevin & Richard. With all those info I should be able to make it but first I need a coffee and a bit of fresh air ;)
Cheers
Seb
- Seb1180_135325Nimbostratus
Feel a bit stupid to ask this but would anyone be so kind to detail me the steps I have to follow on how to achieve this has of course there is no template for dummy and I have tried all I could but doesn t succeed. Thanks a lot
seb
- Kevin_StewartEmployee
The admin guides on the support site can go into much greater detail, but essentially you'd do this:
- Create a pool of web servers for each application
- Create the above iRule and modify as required (change host and pool names accordingly)
- Create a virtual server and apply the iRule (and potentially one default pool)
- Apply an HTTP profile to the VIP config
- Apply any other profiles as required (persistence, SNAT, etc.)
That's really about it. For HTTPS traffic you'll create a separate VIP listening on the same IP address and port 443. All other settings can be the same (HTTP profile, default pool, iRule, etc.). You also need a client SSL profile for this VIP to be able to negotiate SSL with the client. If you choose options 1 or 2 above (wildcard or SAN), you'll need to first acquire one of those types of certificates and private keys, apply them to a single client SSL profile, and then apply that profile to the VIP. If you choose option 3 (SNI) - and remember its SSL TLS and F5 v11 requirement - you would create a separate client SSL profile for each single subject certificate, specify a host name in the "server name" block in the profile that matches the subject name in the certificate, and designate one profile as the "default" (a check box in the profile). Apply all of these client SSL profiles to the VIP.
- Seb1180_135325Nimbostratus
Thanks a lot Kevin. that s basically what I have done but maybe I missed something here or there. Will give it another shot and thanks for helping out the newbie :) I ll have a look also in the admin guide.
Cheers
Seb
Recent Discussions
Related Content
* Getting Started on DevCentral
* Community Guidelines
* Community Terms of Use / EULA
* Community Ranking Explained
* Community Resources
* Contact the DevCentral Team
* Update MFA on account.f5.com