Cannot connect to the Docker daemon (docker.sock / dockerd)

The Docker client talked to a socket or TCP endpoint and nobody answered like a daemon. Either dockerd is stopped, the socket path is wrong, or you are pointing at the wrong host.

How it shows up

docker: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
Cannot connect to the Docker daemon at tcp://localhost:2375. Is the docker daemon running?
error during connect: Get "http://%2Fvar%2Frun%2Fdocker.sock/v1.45/containers/json": dial unix /var/run/docker.sock: connect: no such file or directory

Compose, buildx, and language SDKs surface the same underlying dial failure.

Client vs daemon

1. The client is the docker binary (or library) on your PATH.
2. The daemon is dockerd (Linux) or Docker Desktop’s VM engine.
3. They meet at a socket (/var/run/docker.sock) or a TCP/SSH context (DOCKER_HOST).

This error is “client could not speak to daemon”. It is not a failed docker build inside an already-connected session.

vs permission denied

Frequent causes

• dockerd / Docker Desktop not running — service stopped; Desktop quit; WSL distro not integrated.
• Socket path — rootless Docker uses $XDG_RUNTIME_DIR/docker.sock; snap/podman emulation may differ.
• Bad DOCKER_HOST / context — leftover export DOCKER_HOST=tcp://… pointing at a dead machine; docker context use stuck on remote.
• CI / container-in-container — job expects a sibling daemon or DinD service that was never started; sock not mounted.
• After crash / disk full — daemon failed to start (also check ENOSPC and docker system prune).

Diagnose

# What does the client think it should dial?
docker context ls
echo "DOCKER_HOST=$DOCKER_HOST"
ls -l /var/run/docker.sock 2>/dev/null || echo "no sock"

# Linux service
systemctl status docker --no-pager
journalctl -u docker -n 50 --no-pager

# Can anything accept the sock?
curl -s --unix-socket /var/run/docker.sock http://localhost/_ping && echo

# Desktop / WSL: start Docker Desktop, then
wsl -e docker version

Fix checklist

1. Start the engine — sudo systemctl start docker (and enable if you want it on boot), or launch Docker Desktop and wait until it says running.
2. Clear a bad host — unset DOCKER_HOST; docker context use default.
3. Rootless — start the user service (systemctl --user start docker) and export the rootless socket if documented for your install.
4. Permissions — if the sock exists and you get permission denied, fix group membership (docker.sock permission denied), not “daemon running” alone.
5. CI — ensure the DinD service is healthy and DOCKER_HOST matches (often tcp://docker:2375 or a mounted sock).
6. Space / corruption — if dockerd exits immediately, free disk and inspect logs before retrying pulls.

# After the daemon is up
docker version          # Client + Server sections both populated
docker info | head
docker run --rm hello-world

Related

docker.sock permission denied · docker system prune / reclaim space · ENOSPC no space left · localhost refused to connect · Permission denied (publickey) · getaddrinfo ENOTFOUND

Questions

Published 6 November 2026.