Route to different pools based on /content

Hi Greg,

You can use HTTP classes or an iRule to do this. With HTTP classes, (in the GUI under Profiles : Protocol : HTTP Class) you'd create one class per pool and configure the URI's with string matching in the URI Paths section. Here is a view of the config from the bigip.conf:

profile httpclass qa_class {
   defaults from httpclass
   pool qa_pool
   redirect none
   paths {
      "/QA*"
      "/qa*"
   }
}
profile httpclass uat_class {
   defaults from httpclass
   pool uat_pool
   redirect none
   paths {
      "/UAT*"
      "/uat*"
   }
}

Note that I'm assuming the app isn't case sensitive and you'd want /UAT and /uat to both go to the same pool. You could add a final HTTP class with no filters and specify a pool for all URIs which don't match a prior HTTP class.

Here is an example of an iRule to select between pools based on the URI:

when HTTP_REQUEST {

    Check URI set to lower case with wildcard matching
   switch [string tolower [HTTP::uri]] {
      "uat*" {
         uat_pool
      }
      "qa*" {
         qa_pool
      }
      default_pool {
         qeneral_pool
      }
   }
}

Aaron

Thanks for all the information. That worked great but apparently I was misinformed on part of what they are wanting. They actually want the QA, UAT or Prod removed from the string when it is sent on to the pool. So if you went to https://www.domain.com/QA/Signon.jsp they want it sent to the QA pool but they want the QA removed from the string but keep everything after that.

 

 

Thanks again for all your help.

 

Greg

How about:

HTTP::uri [string range [HTTP::uri] [string first "/" [HTTP::uri] 1] end]


% set uri "/QA/Signon.jsp"
/QA/Signon.jsp
% string range $uri [string first "/" $uri 1] end
/Signon.jsp

Would there be a way to do this to just remove the QA, UAT or Prod from the url without having the full path defined within the iRule. It would be a little cleaner and easier to manage as there could be additional paths that need to get added later on and would rather not have to keep adding rules as they add additional paths.

 

 

Also, is there a best practice as far as how to implement the iRules? When doing multiple things (sending trafifc to different pools based on uri and removing part of the url) do you want to use one iRule or do you just keep them separate?

 

 

Thanks,

 

Greg

 

What do yo mean defining the path? As far as best practices go, there are trade-offs between readability and manageability. If you aren't doing too much, it makes sense to be lumped into one iRule. If there are a crazy number of lines and functions, it might make more sense to split them out.

If, for example, today we had a client going to http://www.domain.com/QA/Signon.jsp and then a month from now development added a second link such as http://www.domain.com/QA/secondpage.jsp I would rather, if possible, to have it set up to send it on to the QA pool as http://www.domain.com/Signon.jsp and http://www.domain.com/secondpage.jsp respectively without having to modify the iRule each time they add a new path. So the iRule would just look for the QA, in this example, to know it needs to send it to the QA pool and then remove the QA from the path before it sends it on. Hopefully that makes a little more sense.

 

 

Thanks,

 

Greg

oh, I think you misunderstood my example. After the first line, I just demonstrate the string command in a native tcl shell. This is the only thing that is actual iRule code:

 

 

HTTP::uri [string range [HTTP::uri] [string first "/" [HTTP::uri] 1] end]

 

 

You'll want to make sure you use this only in the switch cases though so it doesn't arbitrarily pull path information out of all URIs.

I must not be doing something right when trying to tie the two together in one iRule. How do you incorporate the two sections below in to one iRule? When I've tried combining the two it will send it on to the pool but it won't strip off the /QA.

 

 

Section 1:

 

when HTTP_REQUEST {

 

Check URI set to lower case with wildcard matching

 

switch [string tolower [HTTP::uri]] {

 

"uat*" {

 

uat_pool

 

}

 

"qa*" {

 

qa_pool

 

}

 

default_pool {

 

qeneral_pool

 

}

 

}

 

}

 

 

Section 2:

 

HTTP::uri [string range [HTTP::uri] [string first "/" [HTTP::uri] 1] end]

 

 

 

Thanks for your patience,

 

Greg

 

 

You'll have to put section 2 inside of each match in your switch statement, i.e.

"qa*" {
    qa_pool
    HTTP::uri [string range [HTTP::uri] [string first "/" [HTTP::uri] 1] end ]
   }

Thanks Chris. I tried doing that and it did not strip out the /QA, /UAT or /Prod. The url never gets altered.

 

 

Greg