Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

In a load balanced environment: How likely is it for two subsequent calls from a given component to arrive at the same server?

Let me explain: If component A calls component B, which then fails, and Component A retries the call to Component B, how likely is it for the same machine to be accessed for component B?

Let's assume that the software is written to use transactions and the above scenario would all be happening in a single session. I.E. something called Component A, and is still waiting for a reply after Component B is called for the second time.

Also, we are assuming that all calls to all components go via the load balancing service and that all machines in the cluster have all components installed on them.

Effectively, I would like the second call to try a different machine, rather than retrying the same machine. Would I have to do my own manual load balancing? (please say no!!! ;-)

Any help, guidance and feedback would be much appreciated.


share|improve this question
up vote 1 down vote accepted

The answer to this is ENTIRELY based on what's doing your load-balancing.

There's a huge gulf in functionality between the most basic, windows-based software load-balancing through lots of options up to something hugely configurable such as Zeus's ZXTM or Cisco's ACE.

What you're asking for is very easy in any of the higher end solutions, I can't answer for your preferred solution without knowing what it is.

share|improve this answer
Okay, I figured as much. Basically, I needed a default answer, since this depends on the client installation. Could be Windows Loadbalancer, or Cisco or anything else. Getting client to make changes to these setting would probably be hard, so what is the default in most cases (if any)? – Gineer Dec 21 '09 at 14:07
The default answer is "maybe"; in other words, don't rely on it. – womble Dec 21 '09 at 19:26

We successfully used Cisco CSMs (first generation to the ACE) to load balance custom socket calls to specialized backend database servers (running a custom C socket at the time). Previously, the front end web servers would have to time out and retry a different socket on their own, but after deploying to the CSMs, we actually saw a decrease in average request time because the CSMs eliminated all of the timeouts.

The second call "to a different machine" is fairly easily configured on most load balancing hardware. You can usually even configure a "failure" based on the backend server's response.

share|improve this answer

It really depends on what's doing your load balancing and how configurable it is. For example, if you chose to use a Linux box using LVS as your load balancer, you have a choice between several load balancing methods including round-robin (what you're after by the sounds of it), source address hashing etc.

Most other load balancing solutions will have similar functionality as well, so it'll work however you've configured it to.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.