Getting error "docker: command not found" while setting up job

[SohanTirpude]

Select Topic Area

Question

Body

Hello all,

I am trying to run job in container in self-hosted runner but while doing so I am getting "docker: command not found" error while setting up job.

This is content of my workflow yaml file

name: Advanced Workflow

on:
  push:
    branches:
      - main

jobs:
  build-job:
    name: build job
    runs-on: self-hosted
    container:
      image: oven/bun:slim
    outputs:
      build-output: ${{ steps.buildStep.output.buildOut }}
    steps:
      - name: define output variable
        id: buildStep
        run: |
          bunvar=$(bun --version)
          echo "buildOut=$bunvar" >> "$GITHUB_OUTPUT"
  deploy-job:
    name: deploy job
    runs-on: self-hosted
    needs: build-job
    steps:
      - name: test build output
        run: echo ${{ needs.build-job.outputs.build-output }}

And this is the complete output of workflow:

Current runner version: '2.323.0'
Runner name: 'fedora'
Machine name: 'fedora'
GITHUB_TOKEN Permissions
Secret source: Actions
Prepare workflow directory
Prepare all required actions
Error: docker: command not found

Can anyone help me out in fixing this issue?

Thank you.

[brudnak]

You're encountering the error because your self-hosted runner does not have Docker installed, but you're using the container: directive in your workflow, which requires Docker to run containers.

---

✅ Solution: Install Docker on the Self-Hosted Runner

If you're using Fedora, run the following commands on the runner machine:

sudo dnf install -y docker
sudo systemctl enable --now docker
sudo usermod -aG docker $USER

Then log out and log back in, or restart the runner service to apply the group change.

---

🛠️ Alternative: Remove the container: Block

If installing Docker isn't feasible, you can refactor your workflow to run natively on the host without containers. That means removing this part from your workflow:

container:
  image: oven/bun:slim

And instead install bun directly in your job steps using a script or package manager.

---

Let me know if you want help rewriting the workflow without containers!

[SohanTirpude]

Hello @brudnak,

Thank you for the detailed explanation. Well, I don't want to install docker in the self-hosted runner but I do have podman installed and I have set the alias docker for podman.

But it is not working with the alias and I can't afford to install docker on my self-hosted machine. Is there any way to use podman instead of docker?

Kindly let me know if you have any idea on this.

Thank you.

[brudnak]

Hi @SohanTirpude,

Thanks for the update!

While aliasing docker to podman works for many CLI tasks, GitHub Actions' container: directive expects a Docker daemon (dockerd) to manage containers. Podman, being daemonless by default, doesn't fulfill this requirement, leading to compatibility issues.

🔧 Alternative Approaches:

1. Enable Podman's Docker-Compatible Socket:

   Podman offers a Docker-compatible REST API via a socket. You can enable it with:

   systemctl enable --now podman.socket
   export DOCKER_HOST=unix:///run/podman/podman.sock

   However, even with this setup, GitHub Actions may still expect a true Docker daemon and behave unpredictably.

2. Refactor Workflow to Avoid container: Directive:

   Since you have Podman installed, consider removing the container: block from your workflow and install necessary tools directly in the job steps.

   Example:

   steps:
     - uses: actions/checkout@v3
     - name: Install Bun
       run: |
         curl -fsSL https://bun.sh/install | bash
         export PATH="$HOME/.bun/bin:$PATH"
     - name: Run your task
       run: bun install && bun run start

   This approach allows your workflow to run natively on the host without relying on Docker.

[SohanTirpude]

Hello @brudnak,

Thanks for the reply and much valuable information. It seems that with podman I can't run the action. Sad! I didn't know that the container feature was so tightly coupled with Docker.

Anyway, thanks again for the support and info.

Thank you.