shibboleth SP with memcache

[Michael O Holstein]

use repmemcached in place of the standard memcached .. its' available as a diff/patch and it's in the repos for most distributions. the various terracotta (eg: ehcache) stuff also works well for replicated in-memory object cache.

when doing these you generally tell each server to write to all cache servers so it doesn't need to replicate that .. but it'll bootstrap a node that gets bounced.

as Scott just pointed out .. if you can get out of the business of trying to do HA between a bunch of servers it saves a lot of headaches .. just authenticate the user and have owncloud give them a session cookie.

Michael Holstein
Cleveland State University

________________________________________
From: users <users-bounces at shibboleth.net> on behalf of Christian Rost <rost at b1-systems.de>
Sent: Thursday, July 30, 2015 4:13 PM
To: users at shibboleth.net
Subject: shibboleth SP with memcache

Hello,

I stuck at a project here and I am not sure if I understood correctly
how shibboleth works.

Our goal is to install ownCloud on multiple apache Webserver and use the
apache shibboleth module to provide sso-login for the users.

The ownCloud session is prelicated via memcache but I am not sure If
this is also required for shibboleth sessions.

So what happend after one of the webserver crashed and the loadbalancer
switch the user to the second webserver? Are the users still logged in
if the ownCloud session is present on the other node?

Do I need some sort of session replication for shibboleth?

Thanks for your help

Regards,

Christian