Fix “docker daemon is not running” on Linux, macOS, and Windows

Overview

You see this when the Docker CLI cannot reach the Docker daemon (dockerd). Typical message:

• error during connect: this may indicate the docker daemon is not running

The fix is to start the daemon (or Docker Desktop), ensure the socket/pipe is reachable, and correct context/environment misconfiguration.

Table: default daemon endpoints

Quickstart (most cases)

• Linux (systemd):
  1. sudo systemctl start docker
  2. docker info
  3. sudo systemctl enable docker
• macOS (Docker Desktop):
  1. Open Docker Desktop
  2. Wait for “Docker is running”
  3. docker info
• Windows (Docker Desktop + WSL2):
  1. Start Docker Desktop
  2. powershell: Restart-Service com.docker.service
  3. docker info

If it still fails, follow the platform steps below.

Minimal working example

Once the daemon is running, validate with a small container.

Linux/macOS:

# Should print an Engine version and no errors
docker version

# Run a tiny container
docker run --rm busybox:latest echo ok

Windows (PowerShell):

# Engine info
docker version

# Tiny container
docker run --rm mcr.microsoft.com/powershell:nanoserver-ltsc2022 pwsh -c '"ok"'

Expected output includes “Server: Docker Engine” from docker version and prints ok from the test container.

Linux: step-by-step

1. Check service status

sudo systemctl status docker --no-pager

• If inactive: start it

sudo systemctl start docker
sudo systemctl enable docker

1. Inspect logs if it fails to start

sudo journalctl -u docker -b --no-pager

Common log causes:

• Missing cgroups or incompatible cgroup driver.
• Low disk space on /var or / (overlay2 needs free inodes and space).
• Stale lock or socket.

1. Verify the socket

# Should exist and be a socket
ls -l /var/run/docker.sock
sudo ss -xl | grep docker.sock || echo "socket not listening"

If socket exists but dockerd is not running, remove stale socket and restart:

if [ -S /var/run/docker.sock ] && ! pgrep -x dockerd >/dev/null; then 
  sudo rm /var/run/docker.sock
fi
sudo systemctl restart docker

1. Non-systemd systems

# SysV/Upstart fallback
sudo service docker start

1. Permission note If you see “permission denied” instead of “daemon not running”, add your user to the docker group and re-login:

sudo usermod -aG docker "$USER"
newgrp docker

macOS (Docker Desktop)

1. Start Docker Desktop (Applications → Docker or run):

open -a Docker

1. Wait for the whale icon to indicate running, then test:

docker info

1. If it hangs or errors:

• Restart backend

pkill -f com.docker.backend || true
open -a Docker

• Ensure your user has permissions to run Docker Desktop and enough resources (see Performance notes).

Windows (Docker Desktop with WSL2)

1. Ensure Docker Desktop is running. Then:

# Restart Desktop backend
Restart-Service com.docker.service -ErrorAction SilentlyContinue
wsl.exe --list --verbose

1. Restart WSL integration if needed:

wsl.exe --shutdown
Start-Process "C:\Program Files\Docker\Docker\Docker Desktop.exe"

1. Verify the named pipe

Get-ChildItem -Path "\\.\pipe\" | Where-Object { $_.Name -like "docker_engine*" }

If missing, Desktop backend is not up.

1. Windows Server (Docker Engine without Desktop):

Start-Service docker
Get-Service docker

Rootless Docker (Linux)

If you installed rootless mode:

# Start per-user service
systemctl --user start docker
systemctl --user enable docker

# Ensure the client talks to the rootless socket
export DOCKER_HOST=unix:///run/user/$(id -u)/docker.sock

Check:

docker context ls
printenv DOCKER_HOST

Contexts and DOCKER_HOST

A wrong context or stale DOCKER_HOST often causes this error.

• Show contexts:

docker context ls

• Switch to default:

docker context use default

• Clear accidental override:

unset DOCKER_HOST

• For remote engines, ensure the target is reachable (SSH/TCP) and the daemon running on the remote host.

Network, proxy, and firewall

• Corporate proxies or firewalls can block dockerd image pulls; that’s a different error. But if dockerd fails to start due to proxy env in systemd, verify:

systemctl show docker | grep -i Environment

• Adjust or remove HTTP(S)_PROXY in the service drop-in if it prevents startup, then restart.

Diagnostics and logs

• Engine logs (Linux):

sudo journalctl -u docker -b

• Desktop diagnostics (macOS/Windows): Use the “Troubleshoot” option in Docker Desktop, then restart.
• Confirm disk and memory are sufficient:

df -h /
free -h || vm_stat 2>/dev/null

Pitfalls to avoid

• Confusing “daemon not running” with “permission denied.” The former is connectivity; the latter is group/ACLs.
• Leaving DOCKER_HOST set to a dead socket/host.
• Running both rootful and rootless with the same client variables.
• Stale /var/run/docker.sock after a crash; remove it only if dockerd is not running.
• WSL2 distro not set to version 2 or Integration disabled with your distro in Docker Desktop settings.

Performance notes

• Enable auto-start so the daemon is ready after boot:
  • Linux: sudo systemctl enable docker
  • macOS/Windows Desktop: enable “Start Docker Desktop when you log in.”
• Resource sizing (Desktop): allocate enough CPUs/RAM for quick startup; too little RAM can delay or prevent the backend from starting.
• Keep /var/lib/docker on a disk with ample space and inodes; low space can block daemon startup.
• Avoid unnecessary debug logging on production hosts; it can slow startup and fill disks.
• On Linux, prefer overlay2 with an up-to-date kernel for stability.

FAQ

• Why does docker version show only Client info? The daemon is unreachable. Start or fix the daemon as above, then rerun.

• Do I need sudo for docker? Only if your user isn’t in the docker group (rootful). Rootless mode avoids sudo but requires DOCKER_HOST to point to the user socket.

• How do I fix this in CI? Ensure the runner starts dockerd (e.g., systemctl start docker) or use docker-in-docker images that launch dockerd in the job.

• Is Docker Desktop required on Linux? No. Use the Docker Engine package from your distro, manage with systemd.

• Can an outdated kernel cause this? Yes. Very old kernels or incompatible cgroup setups can prevent dockerd from starting; update kernel and ensure cgroup v2 is supported or configure cgroup driver accordingly.