On Slurm clusters, Ghosthub’s SSH host is usually a login node, where interactive agents and long-running processes should not run.
Add an optional Slurm mode that requests an allocation through the login host, then creates or attaches to tmux on the assigned compute node. Reconnecting should reuse the allocation; killing the session should cancel it.
Related to #55, but launch profiles do not track allocations or hand sessions off to compute nodes. Initial scope: Slurm, user-defined resource profiles, and no cluster-specific integration.
On Slurm clusters, Ghosthub’s SSH host is usually a login node, where interactive agents and long-running processes should not run.
Add an optional Slurm mode that requests an allocation through the login host, then creates or attaches to tmux on the assigned compute node. Reconnecting should reuse the allocation; killing the session should cancel it.
Related to #55, but launch profiles do not track allocations or hand sessions off to compute nodes. Initial scope: Slurm, user-defined resource profiles, and no cluster-specific integration.