Skip to content

Choose an execution backend

Choose the API by the runtime and input contract you need. Docker supplies a container engine; Apptainer runs a SIF image on a cluster; Slurm allocates the cluster resources. Installing one does not configure the others.

API Use it for Inputs and runtime Scheduling and resources
RasDocker Wine preprocessing followed by native Linux unsteady computation on Windows or Linux Complete working model, terrain/projection dependencies, and matching published Wine/native images Local Docker CLI; engine must see bind-mount paths. One plan per container, 1–8 solver cores; run_batch is sequential
DockerWorker Native computation in a distributed worker pool Host preprocessing plus an image implementing the worker's run_ras.sh contract Local daemon or remote Docker over SSH; pool configured with cores_total and cores_per_plan
RasPortableDocker Portable steady/unsteady execution requests Complete project bundle, immutable OCI identity, and an image with the portable Python executor; Windows execution on Linux requires a prepared Wine runtime One core per request; execute_pool controls concurrent requests
RasSlurm Schedule portable steady/unsteady execution requests through Apptainer, including Windows HEC-RAS under Wine Same portable request contract, a shared executor SIF, and SlurmSiteConfig One-core contained steps with private node scratch; one exclusive-node pool or a throttled job array; local scheduler or SSH transport
RasApptainer Run a native HEC-RAS 6.x/7.x unsteady plan on Slurm, optionally after a Wine prepare job on the same cluster Solver-ready temporary HDF, boundary and geometry inputs (or a complete project for the prepare job); shared SIFs and a JSON site profile One plan per allocation; solver cores match the allocation; OpenSSH/scp transport
LocalWorker / PsexecWorker Local Windows or distributed Windows execution Installed Windows HEC-RAS and complete projects Local processes or remote Windows hosts

init_ras_worker supports psexec, local, and docker workers. Use RasSlurm or RasApptainer for Slurm execution; the two Slurm APIs have different profiles and receipts.

The Apptainer paths mirror the Docker workflow: RasDocker.preprocess_plan corresponds to RasApptainer.render_prepare_job (Wine preprocessing in an executor SIF), and RasDocker.compute_plan to RasApptainer.render_job (native solve). RasPortableDocker corresponds to RasSlurm.

Start with the appropriate guide

Image compatibility and qualification

The container catalog lists the published image pairs, digests, architecture, source revisions, and qualification scope. These images are tested through RasDocker. Their existence does not establish compatibility with every executor: DockerWorker expects its configured shell runner, while portable execution expects python -m ras_commander.remote.execute_request and a compatible installed runtime.

CLB has run RasSlurm (Wine steady requests) and RasApptainer (native unsteady solves) live on its Slurm cluster with site-built SIFs, and has run Wine preprocessing under Apptainer with the per-task prefix pattern these APIs use. On October 13, 2026 the current APIs were run live with the 6.6 Wine executor built from this repository and the published native 6.6 SIF: a RasSlurm pool and a job array (two concurrent Wine requests each), and a RasApptainer Wine prepare job followed by a native solve, all collected successfully. Qualify the exact images and a representative model on your cluster first. See HPC design and qualification boundaries.

Keep source models separate from working copies. Match the HEC-RAS version across preparation and computation, retain receipts and logs, and inspect the result HDF and modeled time window. A computational success receipt does not establish that the results are acceptable engineering output.

CLB Engineering Corporation  ·  LLM Forward Engineering
RAS Commander Documentation is a free and open-source project maintained by CLB Engineering Corporation. For agencies and firms modernizing H&H workflows with LLM Forward approaches, contact CLB to partner with the engineers who wrote the automation.