> -------- Original Message --------
> From: "Yariv Sadan" < yarivvv@gmail.com>
> Apparently from: erlyaws-list-bounces@lists.sourceforge.net
> To: "Alex Arnon" < alex.arnon@gmail.com>
> Cc: erlyaws-list@lists.sourceforge.net
> Subject: Re: [Erlyaws-list] required steps for clustering yaws ?
> Date: Tue, 18 Jul 2006 09:41:28 -0400
>

[snip]

> >
> > - Use a simple load balancer that round robins requests between web
> > servers, and keep all session data in a replicated Mnesia table. This
> > is a simpler setup in a sense and it's more resilient too. The
> > downside is that the replication may slow things down, but everything
> > has a price :)
> >


This is the option that I'm currently exploring, as it seems to be more generic in nature.
However, there is the problem of expired sessions, specifically the case where a user started a session but for some reason or other never got around to "closing" it.
I think a per-node Session Manager process would be necessary for this, where the connection process (at the start of out()) will notify it that it is using the session, and would manage locking of the session object. To complete the picture, session scrubbing processes should activate periodically and remove sessions that are expired. Distributing this might be done by another table (ordered set) whose key is expiration time.

Thoughts?