flux(1)
Flux is a container-native resource manager and job scheduler for HPC clusters and Kubernetes environments.
Synopsis
flux [OPTION]... COMMAND [ARG]...Description
Flux is a next-generation resource manager designed to manage large-scale computing clusters with improved scalability, flexibility, and ease of use compared to traditional HPC schedulers. It can run on bare-metal HPC systems, within Kubernetes clusters, or in hybrid environments.
The flux command provides a command-line interface to interact with a running Flux instance, allowing users to submit jobs, query status, manage resources, and monitor cluster state. Flux uses a hierarchical architecture where each node can run a Flux broker that communicates with other brokers to coordinate job scheduling and resource allocation.
Common options
| Flag | What it does |
|---|---|
-h, --help | display help message and exit |
-v, --verbose | increase verbosity level (can be repeated) |
-q, --quiet | suppress non-error output |
--version | display Flux version and exit |
-H, --host=HOSTNAME | connect to Flux broker on specified host |
-x, --exit-on-error | exit immediately if a subcommand fails |
Examples
submit a job requesting 4 tasks with 30-minute time limit
flux submit -n 4 -t 30m ./my_job.shlist all jobs in the current user's session with their status
flux jobscancel all running jobs for the current user
flux cancel $(flux jobs -a -n1 -S running -o {id})display available compute resources and their current allocation status
flux resource listshow detailed information about a specific job
flux job info JOBIDallocate 8 nodes for 1 hour and start an interactive shell
flux alloc -n 8 -t 1h bashrun hostname command on all allocated resources
flux exec hostnamedisplay real-time resource usage and job activity
flux top