[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] Lost in IDTOKENS



Hi Carsten,

I'm glad to hear that you're exploring the move from host-based security to IDTOKENS! Your mental model for node <-> auth seems sensible to me!
For users, I think it's perfectly reasonable to stick with FS; it's what 
we do at the CHTC. I think your auto-enrollment solutions effectively 
rely on FS access anyway so security-wise they sound equivalent.
- Brian

On 8/31/23 04:13, Carsten Aulbert wrote:
Hi all,

finally, we are about to move past v8.8 and with going to the 10.7 series on Debian bookworm. During this transition, I would like to get away from the current FS/HOST based security model, especially as the latter is very much frowned upon, but at the moment I'm struggling to get around IDTOKENS and how to pre-create those for users and how to distribute this centrally by a host-based configuration management tool.
My goal is to give every user access to our three local pools, enable 
flocking between them and have condor daemons talking easily between 
each other as necessary, just like they can at the moment. All 
relevant nodes have access to the same shared storage system which 
consists of dozens of different networked file systems, e.g. NFS and 
BeeGFS with Ceph to be added soon.
Currently, I don't see a problem to create a central auth token with 
the pool password and distribute this to every single central 
manager/schedd/execute node. I'm not yet sure, whether to have a 
different central auth token for every pool but even then, 
distributing one or three central secrets is not a big headache - Iâ 
think.
However, at the moment after reading the manual, I don't know how to 
enroll local users into this scheme.
My goal here is that users won't even notice any regression in 
migrating to a different security model, they continue to ssh into the 
schedd machines and can condor_submit/condor_q/condor_rm all they want 
(especially also condor_q -better-analyze which needs direct read 
access to a remote collector).
What I don't want is that users need to start to request tokens which 
we as admins have to manually approve of.
Sure, we could probably write a script which does all this for us, 
generate/approve a token and somehow copies this to the user's 
~/.condor/tokens.d but at the moment I don't really see how Iâ would 
be able to automate this easily[1].
Is there an easy mechanism within HTCondor itself to auto-approve 
locally generated requests (e.g. triggered by the user's ~/.profile or 
first use of a condor command?). From the man page of 
condor_token_request_auto_approve this does not seem to be the tool of 
choice.
Final side note (for now), how would one prohibit a user from 
generating a request for another user's identity? So far, I drew 
blanks from the manual on that.
As always, thanks a lot for some pointers, at the moment I'm 
completely lost between large bits of documentation and lack of 
documentation which could be applied to our set-up.
Cheers

Carsten

PS:â If relevant, I don't expect external submissions from outside our internal networking structure.
[1] Initial problems I see with trying our host based configuration 
management to tackle this includes
* central managers usually don't have knowledge of our Unix users
* schedd hosts know about the Unix users but write access to these users' home directories could prove finicky due to NFS and friends * file servers with local access to the users' files have no local condor installation *â trying to automagically create the proper tokens on the fly does add extra node inter-dependencies which Iâusually try to omit adding
* ...


_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/