Shibboleth Developers

[Chad La Joie <>]

Wilcox, Mark wrote:
> Silly question - what exactly is driving the use case for HA Shib?  And 
> how will you know if you have succeeded?

We have a use case for it here at GU.  We want to use artifacts for some 
things and don't want to rely on a database to store mappings.  So if we 
can sustain system failures and still have everything work okay then 
we'll know we've succeeded.

> For example - do you actually need to replicate state across the nodes 
> in the HA cluster?  If you can avoid state replication, then you don't 
> need this multicast stuff and then clustering becomes a lot easier and 
> more reliable IMHO.

Right, this clustering stuff adds complexity, no argument, and in 
situations that can be addressed by other less complex means I think 
they certainly should be.  In fact, I started writing a Wiki article on 
how you do this (link below).  Some situations can't however, and for 
those, we have this extension.  And as Scott said, in Shib 2.0 I think 
we'll end up were you pretty much need this kind of system or a database 
(which as Scott points out just adds another thing to the mix that could 
fail) if you're going to deploy in a load balanced environment.

https://authdev.it.ohio-state.edu/twiki/bin/view/Shibboleth/LoadBalancedIdP
--
Chad La Joie             315Q St. Mary's Hall
Project Sentinel         202.687.0124