flux-framework / flux-core

core services for the Flux resource management framework
GNU Lesser General Public License v3.0
167 stars 50 forks source link

system instance: add remote job shell capability #2258

Open garlick opened 5 years ago

garlick commented 5 years ago

Provide ability for job-exec module of system instance to launch flux-shell remotely, on a brokerless compute node.

The shell needs to be able to connect back to the system instance to use job-info to obtain R and jobspec; to use barrier to synchronize; and to use kvs to access the guest namespace. Or there needs to be other equivalent mechanisms to accomplish the same thing.

If a Flux instance is launched by these remote shells, the new instance may need a valid FLUX_URI for the enclosing instance to obtain its resource set, advertise a proxy FLUX_URI, etc. (depending on design TBD).

grondo commented 5 years ago

Or there needs to be other equivalent mechanisms to accomplish the same thing.

Recall also the use case of submitting a new job to be run in the parent instance within the batch script of an existing job. This should work as well. (It seems it might be safest to ensure the connector provided on brokerless compute nodes has full functionality)

garlick commented 5 years ago

1281 (tcp connector) may be useful here.

stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had activity for 365 days. It will be closed if no further activity occurs within 14 days. Thank you for your contributions.