Execution Backends ​

Sprocket evaluates a workflow locally and decomposes it into tasks. Sprocket dispatches these tasks to a configured execution backend responsible for starting, monitoring, and responding to task lifecycle events. See the execution model for how a run proceeds end to end.

Sprocket currently supports the following execution backends:

• A local Docker execution backend.
• A remote Task Execution Service backend.
• An LSF + Apptainer HPC backend (experimental).
• A Slurm + Apptainer HPC backend (experimental).

Backend configuration ​

A sprocket.toml configuration file may contain settings for zero or more execution backends, but currently Sprocket uses only one specified backend during a workflow evaluation.

The execution backend is specified in the run section:

[run]
backend = "<name>"

The name corresponds to an entry in the run.backends table. A backend setting is required if there is more than one backend or if the backends table does not contain a default entry.

An example of configuring more than one backend:

[run]
backend = "server1"

[run.backends.server1]
type = "tes"
# ...

[run.backends.server2]
type = "tes"
# ...

If no execution backend is configured in sprocket.toml, the Docker backend is used.

Container protocol support ​

Different execution backends support different container image protocols: Docker accepts docker:// images, TES servers accept docker://, library://, and oras://, and the Apptainer-based backends accept those plus local file:// SIF images. See containers for the full table and the details of each protocol.