sticky session using load balancer

nr673 . at
Thu Dec 17 12:03:16 EST 2015

Thanks Peter. Appreciate your feedback. We want to avoid lengthy
stickyness. Sometimes, the load may not get distributed properly. Let me
try the other clustering options.

On Thu, Dec 17, 2015 at 1:52 AM, Peter Schober <peter.schober at>

> * nr673 . < at> [2015-12-17 01:54]:
> > As mentioned in the following documentation, I enabled the session
> > stickyness for 2 mins in the load balancer.
> >
> >
> >
> > shibboleth2.xml does not declare any StorageService element. By default,
> > Shibboleth would use the memory storage service. I could login into my
> > application successfully. After 3 or 4 minutes idleness, Shibboleth
> looses
> > the session and redirects me to the application welcome page.
> It's not Shibboleth losing the session, it's the loadbalancer losing
> stickiness (as configured) and routing the request to a web server
> that does not have the same Shibbleth session in memory as the one
> that created it initially.
> > Is the clustering possible just by enabling the session stickyness
> > in load balancer?
> No.
> With permanent stickiness you could make something like that work,
> until one of the 4 machines goes down, then everyone having a session
> on that SP would have to log in again.
> Many would still find that a workable compromise, I suppose.
> > Or do I need to use shared process or shared database approach?
> Yes (modulo the compromise mentioned above). If you can't make sure
> requests will always end up on the same server then you need to make
> sessions available to all the servers where requests might end up.
> I.e., you can't keep them in memory on one server and expect things to
> magically work if you route requests to another server.
> -peter
> --
> To unsubscribe from this list send an email to
> users-unsubscribe at
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the users mailing list