Announce: Database Backed Storage Service

[Paul Hethmon]

Keith,

I've got it in production for about 2 dozen customer installations,
ranging from a user population of a 1k to 90k. So I definitely feel its
ready. Terracotta for me was a nightmare to tune. I then tried using
Infinispan and that worked better, but still had occasional problems on my
network and I had some customer hosted installations that it never would
work on. This has been very solid for us.

Just be aware that it is designed to fail silently. So if the DB layer
fails for any reason, the storage service will default to a local storage.
So a user may have to login again when they shouldn't have to, but they
won't be denied access if the DB fails.

cheers,

Paul


On 9/17/13 10:55 AM, "Wessel, Keith" <kwessel at illinois.edu> wrote:

>Paul,
>
>We're considering using your storage service as one of the options for
>replacing Terracotta. Are you using this in production yourself yet?
>We'll test it heavily before putting it in production, of course, but do
>you at least feel it's ready for production use?
>
>Thanks,
>Keith
>
>
>From: users-bounces at shibboleth.net [mailto:users-bounces at shibboleth.net]
>On Behalf Of Paul Hethmon
>Sent: Tuesday, August 27, 2013 9:30 AM
>To: Shibboleth Users
>Subject: Announce: Database Backed Storage Service
>
>I finally got my database backed storage service packaged up for public
>release:
>
>https://wiki.shibboleth.net/confluence/display/SHIB2/Database+Backed+Stora
>ge+Service
>
>That page hopefully gives all the information needed to get it running. A
>few key points:
>
>1. It uses Hibernate for the data access layer, but I have only connected
>it to MySQL.
>2. The build script is Ant based which means dependencies will be a pain.
>I do have a goal to switch it to Maven, but don't hold your breath.
>3. There is a binary distribution in the Subversion repository. Java 6.
>4. I run a lot of copies of the IdP in my job that are logically
>distinct. You will see that the storage service requires a constructor
>argument to let it know which bucket to place the sessions. What you use
>does not matter. Only sessions with the same key are considered to be in
>the same logical group.
>5. The code will hide ALL database failures. The idea was that a database
>failure would not prevent a user from logging in. It might require them
>to re-authenticate, but they will be able to login.
>
>Please let me know what I forgot in the documentation, either on this
>list or direct is fine.
>
>thanks,
>
>Paul
>
>
>
>--
>To unsubscribe from this list send an email to
>users-unsubscribe at shibboleth.net