How to Generate Visitor ID/UUID

You could incorpate something like the following into your irule

 

 

set unique [md5 [IP::client_addr]]

 

HTTP::cookie insert name "UUID" value $::unique

 

 

Then you could base your persistance off of what is in the cookie or you can insert into the HTTP headers.

 

 

Hope this helps

 

 

CB

 

 

 

Unfortunately, with super proxies and corporate proxies the client IP wouldn't be enough as MD5 isn't collision resistant. But, I'll have to think a bit about how collision non-resistant MD5 is.. It might be good enough.

[EDIT: double post]

Anybody know what the cookie that the BigIP drops is exactly?

 

 

i.e. BIGip=3859747244.16671.0000

 

 

Is it pretty unique, and is it accessible within an irule? I was thinking it must be as my assumption is it's used for the persistence table.

If not, my thought was to do something similar to what tcllib uuid is doing:

 

 

Append the following together:

 

clock seconds Timestamp

 

clock clicks system incrementing counter

 

pid additional entropy

 

server_addr get some spatial uniqueness

 

client_addr add some more spatial uniqueness

 

 

Then md5 this all, and convert the result to a UUID format.

The BIG-IP persistence cookie is an encoded (not encrypted) form of the pool member IP and port. The cookie name by default is BIGipServer. The encoding is described in SOL6917 (Click here). This codeshare example (Click here) shows how to decode it in an iRule.

 

 

Are you trying to generate a unique ID for handling sessions or for persistence? If the former, I could see why you'd want to prevent collisions. If the latter, I'm not so clear. Can you explain why you would need a globally unique token for persistence? Why couldn't you just the default cookie insert persistence with a expiration time set according to your requirements? I don't understand why the persistence cookie would need to be globally unique if you're just trying to give the client a token with which you get them back to the same pool member.

 

 

Aaron

Thanks, on the BigIP cookie. I was hoping it might be a unique id, bt alas not.

 

 

What we're doing is creating a unique visitor ID across all our clusters that is used in our applications. We have several clusters and no common, unique id that can be shared to correlate across products. To solve it in the applications can be error prone from a development perspective, and we want to drop a unique visitor cookie at the load balancer. Most of the products go through the load balancer, so it's a good place to do it.

 

 

Right now I'm moving forward with implementing portions of the tcllib UUID to get as unique as I can identifier. My main concern now is performance and load on the LB.

I ended up doing the following below... I'm not Tcl master, and some of the binary scan/format was simply copied from the tcllib.

My only thought is, does md5 use the rand function, and does the rand function use clock. If so, then either clock or rand are redundant and probably not increasing entropy.

when RULE_INIT { 
 set ::cn "lbId" 
 set ::ce 15552000 
 } 
  
 when HTTP_REQUEST { 
 if { not ([HTTP::uri] starts_with "/static" or [HTTP::uri] starts_with "/assets") } { 
 if { not [HTTP::cookie exists $::cookieName] } { 
 append s [clock seconds] [IP::local_addr] [IP::client_addr] [expr { int(100000000 * rand()) }] [clock clicks] 
  
 set s [md5 $s] 
  
 binary scan $s c* s 
 lset s 8 [expr {([lindex $s 8] & 0x7F) | 0x40}] 
 lset s 6 [expr {([lindex $s 6] & 0x0F) | 0x40}] 
 set s [binary format c* $s] 
 binary scan $s H* s 
  
 append u [substr $s 0 8] "-" [substr $s 8 4] "-" [substr $s 12 4] "-" [substr $s 16 4] "-" [substr $s 20 12] 
 unset s 
  
 HTTP::cookie insert name $::cn value $u 
 } else { 
 set u [HTTP::cookie value $::cn] 
 } 
 set h [HTTP::host] 
 } 
 } 
  
 when HTTP_RESPONSE { 
 if { not ($u eq "") } { 
 set h [findstr [getfield $h : 1 ] ".eharmony."] 
 if { $h eq "" } { 
 HTTP::cookie insert name $::cn value $u path "/" 
 } else { 
 HTTP::cookie insert name $::cn value $u path "/" domain $h 
 } 
 unset h 
 unset u 
 HTTP::cookie expires $::cn $::ce 
 } 
 }

MD5 is a deterministic algorithm and does not use rand.

 

Rand does not use clock. It only uses its seed (which most people set with clock).

 

 

As for performance, it turns out that

 

HTTP::cookie and TCL math are your most expensive operations.

Here's one method we tried:

 

 

How random is rand()?

 

http://devcentral.f5.com/Default.aspx?tabid=53&forumid=5&tpage=1&view=topic&postid=813535

 

 

Aaron