sparkrun stop

sparkrun stop [target] [options]

Stop containers for a running inference workload across all hosts. TARGET can be a recipe name or a cluster ID (from sparkrun status output). When --all is used, TARGET is optional and all sparkrun containers will be stopped.

Options

Option	Description
--all / -a	Stop every sparkrun workload on the cluster. No target needed.
--hosts / -H	Comma-separated host list
--hosts-file	File with hosts
--cluster	Use a saved cluster by name
--tp / --tensor-parallel	Match host trimming from run
--port	Override port (to match run-time override)
--served-model-name	Override served model name (to match run-time override)
--dry-run / -n	Show what would be done

Examples

# Stop a specific recipe
sparkrun stop qwen3-1.7b-vllm
sparkrun stop nemotron3-nano-30b-nvfp4-vllm --tp 2
sparkrun stop nemotron3-nano-30b-nvfp4-vllm --cluster mylab

# Stop all sparkrun containers on a cluster
sparkrun stop --all --cluster mylab
sparkrun stop --all --hosts 10.24.11.13,10.24.11.14

Caution

The --hosts/--cluster/--tp flags must match the original run command so sparkrun can find the right containers. If you launched with --tp 2, pass --tp 2 to stop as well.

Discovery and reporting in 0.3.0

--all discovers workloads through the same cross-executor status source as sparkrun status, so it also stops workloads launched by the native local executor — not just Docker containers.

Teardown now verifies and counts what it actually removed instead of trusting the removal command’s exit code, so a stop that fails on one host reports that host rather than exiting 0 while leaving VRAM held.