Shibboleth Developers

[giacomo tenaglia <>]

[I Cc-ed the mail to shibboleth-users ML, I think it's the best place
for this subject.]

On Thu, Nov 22, 2007 at 05:33:49AM -0500, 

 wrote:
> Sorry to disturb you, I try to found some information about SP cluster. But 
> don’t really found correct answer.
> 
> Some documentation had read show me that clustering SP is very simple. I 
> don’t think so.
> 
> Shibbolleth Session cache seems to be a hardest problem.

Hi,
yes the session cache is the hardest problem:
https://spaces.internet2.edu/display/SHIB/CPPSPLoadBalanced

> I don’t find good information for clustering SP.
> 
> Use an embedded MySQL session cache? Not really good, embedded mysql cannot 
> be replicated…
> 
> Could you help me to find a way to search …

If you can't or don't want to modify your application to use Shibboleth
only for the first login phase, you have to go for a shared cache.

I've tried to use one shibd shared between cluster machines and using a
shared TCP socket, but it's very slow.
I think you can:
- use the embedded MySQL plugin, and share the data directories between
  cluster machines, but you have to check if it doesn't mess up the db;
- use the remote MySQL session cache plugin developed at Stanford,
  waiting to be included in the upstream C++ SP, but already present in
  the Debian packages: https://bugs.internet2.edu/jira/browse/SSPCPP-33
  I'm doing some tests with it and I think it's the best solution.

giacomo

-- 
giacomo tenaglia
Technical student at CERN IT-DES/SIS
Phone +41 76 5003376 - 
sip: