Shell into a running service

Need to peek at a file, check an environment variable, or run a one-off task inside a live container? bootload gives you an interactive shell and one-off command exec straight into a running replica — no SSH keys, no sidecar, no re-deploy. It's an interactive PTY bridged over the API to the replica's node.

Open a shell

bootload shell <service>

This drops you into an interactive shell inside one running replica of the service — resize-aware, just like a local terminal. Exit it like any shell (exit or Ctrl-D).

Run a one-off command

For a single command (and to script it), use exec with everything after -- as the command line:

bootload exec <service> -- env
bootload exec <service> -- cat /etc/resolv.conf
bootload exec <service> -- ./manage.py migrate

What you need

• A running replica. Exec attaches to a live replica, so the service must be scaled to at least one. If it's anchored at zero, scale it up first: bootload scale <service> --replicas 1.

• Write access. Your role in the project must be developer or above, and your API token needs the services:exec scope.

• Confirmation for production. A service marked production requires an explicit --confirm flag, so you don't poke a live workload by reflex:

  bootload shell <service> --confirm
  

Every exec session is recorded in the audit log (who, which replica, the command).

Good to know

• You land in one replica. A service with several replicas will connect you to whichever one is currently serving — don't assume it's the same one twice.
• Changes to the container filesystem are ephemeral: anything you write outside a mounted volume is gone when the replica is replaced (a redeploy, scale, or reschedule). Use it to inspect and debug, not to make lasting changes — bake those into your image or put the data on a volume.
• The shell runs as whatever user your image's entrypoint runs as, with the same environment (including your project secrets) the workload sees.