Docker Executor

The Docker executor is the default execution mode for local development and single-host deployments. It uses the Docker daemon on the host machine to run job steps as individual containers.

When to Use

• Local development and testing
• Single-machine CI pipelines
• Demos and evaluation environments

How It Works

When the Scheduler receives a job in Docker mode:

1. For each step, it pulls the specified Docker image (if not already cached).
2. It creates a container from that image, injecting the step's script and environment variables.
3. A shared Docker volume is mounted at /shared across all steps of the same job.
4. Steps are executed sequentially. The exit code of each container determines success or failure.
5. Logs are collected from the container's stdout/stderr and forwarded to the log manager.

Configuration

Set the following environment variable in your .env file (or compose.yml):

HADES_EXECUTOR=docker

This is the default value, so no changes are needed when using the provided compose.yml.

Resource Requirements

Requirement	Detail
Docker socket	Mounted at /var/run/docker.sock — the Scheduler must have access to it.
Network	The Scheduler and all job containers share the hades Docker network.
Volume	A named or anonymous volume is created per job for the /shared directory.

Limitations

• Single host only — all containers run on the machine where the Scheduler is running.
• No auto-retry — failed steps are not automatically retried; the job is marked as failed.
• No resource quotas — containers are not constrained by CPU or memory limits by default.

For production workloads requiring scalability, retry logic, and RBAC, use the Hades Operator instead.