maand run_command

run_command runs an arbitrary shell command on one or more workers over SSH from the maand CLI host.

CLI

maand run_command [command] [flags]

Examples:

maand run_command "uptime"
maand run_command "df -h /opt/worker" --workers 10.0.0.1,10.0.0.2
maand run_command "hostname" --concurrency 4
maand run_command "uptime" --ignore-failure

Exit code

Exit 0 when every selected worker succeeds. Exit 1 when any worker fails (including with --ignore-failure, which still runs reachable workers and writes their logs before reporting failures).

Prerequisites

• Initialized bucket with workers in workspace/workers.json and a prior maand build.
• SSH key at secrets/<ssh_key> (from maand.conf) authorized on workers.
• Host tools: bash, ssh (checked before SSH).
• Target workers must have bash and timeout on PATH (sudo when use_sudo = true).

Notes

• Commands run on workers as the configured ssh_user (default agent).
• Working directory: scripts start in the worker's bucket directory /opt/worker/$BUCKET_ID, so relative paths like jobs/<job>/... resolve directly. If the bucket has not been deployed yet (no such directory), it falls back to the SSH login directory so host-bootstrap commands still work.
• BUCKET_ID and WORKER_IP are exported into every script.
• Concurrency: without --health_check, at most -c workers run at once; the next worker starts as soon as a slot frees (command output streams to the terminal as each worker runs). With --health_check, workers run in fixed batches of -c and jobs are health-checked after each batch.
• With --ignore-failure, worker command output still streams for reachable workers; each failure is printed to stderr as it happens, and failures are reported again at exit. --ignore-failure and --health_check cannot be used on the same command.
• This is separate from hooks (pre_deploy, job_control, etc.) which use the runtime API and job workspaces.