4:19 p.m.
I think the only thing that doesn't scale very well as it pertains to
number of clients is logout. Logout for OIDC requires a redirect uri.
We validate this uri by iterating over every client's register register
uri patterns.
We don't have any services on key rotation. That's all something you'd
have to implement yourself.
On 5/16/16 3:37 PM, Henryk Konsek wrote:
IMHO mapping device per use should be fine. KeyCloak scales well even
for hundred of thousands of users, so it will handle gazillion of
devices as well.
Cheers!
pon., 16.05.2016 o 16:29 użytkownik Aikeaguinea
<aikeaguinea(a)xsmail.com <mailto:aikeaguinea@xsmail.com>> napisał:
This is disappointing news, as when I asked this same question
back in January the answer was that the intention is to have
Keycloak scale to hundreds if not thousands of clients, and if
there were issues you'd work with us on that.
There's more to this issue than having a custom authenticator; the
client interface allows you to click one button and generate the
jks file containing the client's private key. We would need this
not only for the first time a device is set up, but for key
rotation on an ongoing basis.
Are there ways to plug into the user management interface to allow
generation of non-username/password credentials for a user?
On Fri, May 13, 2016, at 02:11 AM, Stian Thorgersen wrote:
> Hi,
> That's a very interesting use-case. One which we have wanted to
> look into ourselves, but haven't had the resources. Ideally I'd
> say we'd have a device concept in Keycloak as they're not
> strictly clients or users. They'd most likely be backed by users,
> but would have different screens for configuration and would have
> separate authentication flows. That would require a fair bit of
> work to add though.
> In the mean time I don't think clients are a good fit as Keycloak
> is not currently designed to have large amounts of clients, both
> for manageability and performance. Both of the issues can be
> overcome fairly easily, but that would require some work.
> The best solution in my opinion is to use users and implement
> your own custom authenticator to handle IOT devices. It's fairly
> simply to do and gives you the ability to handle authentication
> of the devices exactly how you want to. See
> http://keycloak.github.io/docs/userguide/keycloak-server/html/auth_spi.html
> for more details.
> I'd appreciate if you kept me updated on your progress as I'm
> very interested :)
> On 12 May 2016 at 10:29, Matuszak, Eduard
> <eduard.matuszak(a)atos.net <mailto:eduard.matuszak@atos.net>> wrote:
>
> Hello
>
> We are planning to get a lot of devices, identifyable by
> individual certificates, into an IOT-system being designed
> and developed at the moment. We choosed to authenticate all
> actors (users, software components and devices as well) by
> OIDC-tokens and (pre)decided to use Keycloak as ID provider.
> User and software components are quite straightforward to
> handle with Keycloak (as Keycloak users with the help of a
> user federation provider & id brokerage and for applications
> as Keycloak clients respectively). But I am not sure of how
> to represent our devices (we want to support hundreds of
> thousands of them later on!) by Keycloak means.
>
> It seems that we essentially have 2 possiblities to register
> a device in Keycloak
>
> * As a user
> * As a client
>
>
> By representing devices as Keycloak clients we might take
> advantage of the ServiceAccount (Oauth-Client Credential)
> flow and become able to implement it via (dynamic!)
> registration and it and seems, that we will even be able to
> authenticate our device by their certificates by choosing
> "Signed Jwt" as authenticator option.
>
> My question is, if it would be a good idea to register a very
> big amount of devices as Keycloak clients with regards to
> performance and manageability. In principle I would prefer a
> user-representation (faciliting usage of user federation
> provider & id brokerage for instance), but as far as I
> understood, the appropriate flow would be Direct Access
> (ResourceOwnerPassword Credentials) and here we can only deal
> with username/password instead of certificates.
>
> Do you have any suggestions or hints (even the conclusion,
> that Keycloak is not the suitable ID-provider-implementation
> for large-scale IOT-systems)?
>
> Best regards, Eduard Matuszak
>
>
>
> _______________________________________________
> keycloak-user mailing list
> keycloak-user(a)lists.jboss.org
> <mailto:keycloak-user@lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/keycloak-user
>
> _________________________________________________
> keycloak-user mailing list
> keycloak-user(a)lists.jboss.org <mailto:keycloak-user@lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/keycloak-user
--
Aikeaguinea
aikeaguinea(a)xsmail.com <mailto:aikeaguinea@xsmail.com>
--
http://www.fastmail.com - Access your email from home and the web
_______________________________________________
keycloak-user mailing list
keycloak-user(a)lists.jboss.org <mailto:keycloak-user@lists.jboss.org>
https://lists.jboss.org/mailman/listinfo/keycloak-user
--
Henryk Konsek
https://linkedin.com/in/hekonsek
_______________________________________________
keycloak-user mailing list
keycloak-user(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/keycloak-user