High Avaibility

[Phil Mayers]

On 01/03/12 10:16, Anto wrote:
> Hello
>
> In the coming days I will set up a freeradius server for access
> control and accounting. I've been looking for information on
> freeradius and high availability, since my idea is to have two servers
> in case one fails, continue to operate with the other, but I just
> found information. So I turn to the list, in case I can recommend
> someone with experience on stage.
>
> I do not know if it is feasible to have a server as master and one
> slave, when the main falls, the other up the interface. If there is
> some kind of balancer radius and use both servers, etc..

This is a very vague question. You're going to get a lot of either 
too-vague or too-specific answers.

A few things you need to specify:

  1. When you say "high availability" what are you hoping to achieve?
  2. How long can you tolerate when an unscheduled outage for? 1 second 
or 60?
  3. Do your RADIUS servers talk to external data sources (SQL, LDAP)?
  4. Do you care about load-balancing, or just high-availability?

I'll make a few comments:

Most NASes support 2 (or more) RADIUS servers, and will fail over when 
they detect an outage. For resilience, you just need to build two RADIUS 
servers on different IPs, and specify these in your NAS.

You don't need a load-balancer or other complications, and they will 
just make things less reliable.

Making "redundant" RADIUS servers is easy; you just build two machines, 
and run FreeRADIUS on each with the same config. The "hard" bit is 
replicating any data sources between them (LDAP, SQL) and handling 
"writes" such as accounting packets into SQL, SQL session counters, and 
so on.

You need to be more specific about what you're doing and what you want 
to achieve.