If you're attending the HTCondor Workshop this week (registration is free & still open), please join my session on "From Identity-Based Authorization to Capabilities: SciTokens, JWTs, and OAuth" on Friday at 11am ET (https://indico.cern.ch/event/936993/contributions/4020913/) followed by Jason and Zach presenting on "Allow HTCondor jobs to securely access services via OAuth token workflow" (https://indico.cern.ch/event/936993/contributions/4022103/). Plus there's other interesting sessions all week!
-Jim
> Unfortunately, I was unable to attend the sessions, but I hope they went
> well.
Thanks!
> One topic that might have turned up is about how jobs running within
> HTCondor might access services.
>
> Aside from Jason and Zach's work, one option might be to provide the job
> with the SciToken (if any) used to create the job. I believe this is
> (roughly) what is shown on slide 17 of your presentation, with the job
> interacting with a data server.
We didn't have time for an HTCondor SciTokens demo at the workshop last week. Jason's presentation (https://indico.cern.ch/event/936993/contributions/4022103/) focused on the example of submitting a job that accesses box.com storage, which demonstrates the HTCondor OAuth mechanism that we also use with SciTokens. For prior demos we've shown HTCondor jobs accessing StashCache, XrootD, and CVMFS storage.
> I was wondering if you would be interested in including dCache as a demo
> service that supports SciTokens?
Yes! I'm aware of the dCache-SciTokens interop work, but I've neglected to put anything about it on https://scitokens.org/. Is https://dcache.org/old/manuals/UserGuide-6.0/webdav.shtml a good reference on dCache's SciTokens support that I could link to?
> Currently, there are a few dCache instances that trust the
> "https://scitokens.org/dteam" issuer to identify members of the DTEAM VO.
>
> I believe it would be possible to extend this to support for some other
> SciToken issuer (perhaps "https://demo.scitokens.org") on
> prometheus.desy.de. This would allow easy demonstrations of this
> job-storage interaction working with real-world services.
Yes, that'd be great -- so people would be able to get a token from https://demo.scitokens.org and read something from prometheus.desy.de. Presumably you wouldn't want to allow writes, since there's no restriction on who can get tokens from https://demo.scitokens.org.
https://demo.scitokens.org/.well-known/openid-configuration should contain what you need to allow tokens from the https://demo.scitokens.org issuer. We could add a link to https://scitokens.org and https://demo.scitokens.org with instructions on using the token with prometheus.desy.de (i.e., what the token payload should be set to). Maybe add another button for "set payload for access to dCache demo".
The source for the https://demo.scitokens.org site is at https://github.com/scitokens/scitokens-heroku. While we're at it, we should update the demo to match the current SciTokens specification (i.e., "scope" instead of "scp").
Once that's all working, we could work on a dCache-SciTokens demo that also involves HTCondor.
Regards,
Jim
> So, I think the WebDAV link is fine for now (thanks), but I hope we can
> provide a better link in the future.
OK I've added the dCache link on https://scitokens.org/.
> I think allowing write access would be no problem if the OP was backed
> by, say, CI-Login based authentication.
Sounds good. We're very close to adding SciTokens issuance on https://demo.cilogon.org/.
> So, would it be possible to update demo.scitokens.org so that is uses
> the "scope" claim by default?
Yes. I submitted a pull request for Derek's review:
https://github.com/scitokens/scitokens-heroku/pull/3
Regards,
Jim
--
You received this message because you are subscribed to the Google Groups "SciTokens Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss+u...@scitokens.org.
To view this discussion on the web visit https://groups.google.com/a/scitokens.org/d/msgid/discuss/ad27d148-f080-3a23-2ca3-9c6c1555748d%40desy.de.