Re: [Condor-users] Queueing ssh sessions?

[Ian Chesal <ICHESAL@xxxxxxxxxx>]

> :I can tell you know: this is a nightmare support situation.
> Especially
> :if you've got a lot of different tools. But particuarly with MatLab
>
> Amen to that!

I'm on a tear with all our tool vendors to simplify license consumption
models. Some are listening, others don't care. I'd really like to get
the FlexLM people in a room and work them over for a bit as well. That
they don't let you control, interactiviy, how feature lines are consumed
is simply...backwards. It's an antiquated technology who's time has come
and gone I think. If you could set env vars that say "this shell can
only access this feature line" it would all be pretty sane to manage.

> Currently here the Matlab licenses float and we have enough so that
> works out (usually :)  The Parallel Computing Toolkit license is the
> only one that's special to the cluster since in this model the worker
> nodes only need the PCT license not a regular Matlab license or any
> other toolkits, so for us this is simple.
>
> The new interest in actual parallel jobs is the current complicating
> factor.  Unlike a "regular" parallel universe job the master "task0"
> job is called interactively and spawns the worker nodes when asked.  I
> haven't yet been able to map this to a condor submission model (though
> this discussion is getting me closer looking at how you queue
> interactive jobs).

Users tell our system: I want this type of interactive session and put
tool A, B and C in it for me. Our system spawns a job that starts the
session with the tool binaries in the path and the license settings in
their environment. That's the only way users get access to tools around
here. We don't keep licensed tools or license settings in their
environments by default. They always use the system to get tools.

> If I let people go around the queue I get license scheduling hell, but
> saying sorry you can't do that isn't any better.

Yea, corporate mandates can be very handy things for making users do
your bidding. I wouldn't even want to try this approach in academia. Too
many outliers there when it comes to operating methods. ;)

- Ian

Confidentiality Notice.
This message may contain information that is confidential or otherwise protected from disclosure. If you are not the intended recipient, you are hereby notified that any use, disclosure, dissemination, distribution,  or copying  of this message, or any attachments, is strictly prohibited.  If you have received this message in error, please advise the sender by reply e-mail, and delete the message and any attachments.  Thank you.