CSM - HTTP1.1 and persistent rebalance

Unanswered Question
Nov 23rd, 2007


I have a vserver configured with a SLB policy and persistent rebalance. The policy uses a sticky based on JSESSIONID cookie. If my clients HTTP traffic comes in HTTP1.0, I have no problems; if it comes in HTTP1.1 I have problems. Sniffing the traffic, I see no such cookie and several packet retransmissions with HTTP1.1. So, my questions are:

- If I'm doing persistence based on that cookie and if no cookie is present, what persistence does the CSM?

- If I have persistence rebalance configured, why am I having problems with HTTP1.1 as it is supposed to "help" HTTP1.1 flows?


I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Gilles Dufour Fri, 11/23/2007 - 05:55

if there is no such cookie, there is no stickyness. The connection is simply balanced.

The 'persistence rebalance' command is never a way to help. It forces the CSM to inspect all packets of a connection and rebalance if a new cookie appears.

You'll need to share the trace if you want us to view what is the problem exactly.


joao.lopes Fri, 11/23/2007 - 06:19

Thanks for your reply.

Before sharing the trace, can you please explain how "the connection is simply balanced", mainly for HTTP1.1?

Even without the cookie presence at HTTP packets, I get polity matches on CSM (default policy scores 0 matches). Why?

Relative to persistence rebalance, even for HTTP1.1, you advice not to use? Do you think my problems are caused by the use of this command (remember that for HTTP1.0 and have no problems)?


Gilles Dufour Fri, 11/23/2007 - 06:58

if your policy as no match statement, all traffic will match it and the rule is then applied - sticky based on cookie and if no cookie just loadbalance.

I do not have the trace, so I first do not know what is your problem exactly and second can't tell you if the command is causing it.

Is your browser using pipelined requests ? [this is a feature of http1.1] Because this is not supported by the CSM and it will create problem if you have 'persistence rebalance'.

This command is necessary if you have a proxy coming in sharing the connection between multiple clients all with a different cookie.


joao.lopes Fri, 11/23/2007 - 08:01

Can you send me your email in order to fwd the capture to you?



This Discussion