You tell us ;)

I had in mind a while back that we'd support storing "configuration" such as realm settings in a config file. Then we'd store only clients, users, roles, etc.. in the database. How realistic it is that we'd implement it is unsure though.

On 2 March 2016 at 14:15, Edgar Vonk - Info.nl <Edgar@info.nl> wrote:
Hi Stian,

I realise that. However so far it is working ok for us. We do the re-importing because we really want a continuous delivery pipeline and doing run-time configuration is very different from that. We really only perform run-time configuration to develop and test stuff, immediately after which we export it so that we can update the resulting realm JSON in Git. We work in a similar fashion with other products like content management systems where we make a very clear distinction between configuration (=source code) and run-time managed content. Our CMS of choice (Magnolia) offers a nice upgrade mechanism that we use to re-import all our configuration on every upgrade.

No, we do not re-import users. We see users (just like user groups, role mappings, group mappings and of course sessions) as run-time data (‘content’ to make the link to a CMS). 

In our set-up we store all this run-time data in Active Directory (well, except the sessions, which we do not store persistently yet) and therefore this re-importing in the Keycloak database works for us (to some degree).

A question for you/the Keycloak team: how do you see Keycloak fit into a fully automated continuous delivery pipeline? Most configuration changes in Keycloak are so fundamental to the set-up that I cannot see that you would want to change these on the fly on a production environment at all. It would be very easy to break stuff in my opinion.

cheers

Edgar


On 02 Mar 2016, at 14:00, Stian Thorgersen <sthorger@redhat.com> wrote:

Re-importing everything on each startup is not really something we're supporting. Keycloak wasn't really designed for that and the focus is more on run-time configuration. Do you re-import users as well?

On 2 March 2016 at 13:39, Edgar Vonk - Info.nl <Edgar@info.nl> wrote:
Thanks Stian!

We will have a look at both options.

Concerning clustering we have a different challenge which is that we currently re-import all Keycloak realm data on every start up of Keycloak (and because we do continuous delivery and are developing actively this is multiple times a day). This because we treat all (realm) configuration as source code for which our Git repo is leading.

Effectively this means that we recreate the Keycloak database for every new deployment and of course a cluster is not going to help us here when it comes to uptime. Not sure how to deal with this as yet. Ideally we would want some sort of realm update/patch mechanism instead of a full import but that sounds rather complex to implement.

cheers

On 02 Mar 2016, at 13:23, Stian Thorgersen <sthorger@redhat.com> wrote:

The tokens themselves are not stored, but can be verified by Keycloak as long as the user session is active. So your question is how to make user sessions persisted. We do not support persisting user sessions at the moment. You have two choices:

1. Add an additional node and configure set owners to 2 for the user session caches, or change it to a replicated cache. See the clustering section in the docs for more details.
2. Try to configure Infinispan to persist the sessions. See https://docs.jboss.org/author/display/WFLY10/Infinispan+Subsystem for more details.

On 1 March 2016 at 20:56, Edgar Vonk - Info.nl <Edgar@info.nl> wrote:
Hi all,

What would we need to do to make Keycloak user sessions persistent in the database?

I think the information in: http://lists.jboss.org/pipermail/keycloak-user/2015-April/001921.html is not relevant anymore with Keycloak 1.9.0? Specifically:

"userSessions": {
        "provider": "jpa"
    }

Does not seem to work (“Failed to find provider jpa for userSessions”). User sessions are now managed using Infinispan by default if I understand correctly: http://keycloak.github.io/docs/userguide/keycloak-server/html/clustering.html#d4e3292 ?

Is there a way to make user sessions persistent? 

Our issue is that we send out a lot of activation (‘update password’) emails from our (single) Keycloak server to new users and since we have a continuous delivery pipeline Keycloak does down and up quite a bit and every time it restarts all temporary log in tokens used for these update password actions are lost (since they are stored in memory only). And if I understand correctly these tokens are actually a sort of user sessions.

cheers

Edgar


On 29 Feb 2016, at 17:52, Edgar Vonk - Info.nl <Edgar@info.nl> wrote:

Hi,

See if I understand this correctly: in the default set up of Keycloak sessions and temporary tokens are not persisted in the Keycloak database? So consider this scenario:

1/ login as admin to master realm
2/ go to Users - Credentials and send a ‘Update Password’ reset action email
3/ user receives an email with a link with a unique token to update his/her password in Keycloak
4/ Keycloak server is restarted for whatever reason
5/ the temporary ‘login action token’ no longer exists and the link from 3/ no longer works

Is this correct and expected behaviour?

And if so, can somebody maybe point us in the direction to solve this? I.e. by making sessions/tokens by persistent I guess.

cheers

Edgar


_______________________________________________
keycloak-user mailing list
keycloak-user@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/keycloak-user