roadmap
- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
Prepare for Docker interview questions on how images are built and how containers run, ship and break. Covers images and registries, networking and storage, security, Compose and Swarm.
379 of 379 questions in your plan: 0 known, 0 in progress, 0 due, 0 skipped, 379 new. 0 cut, outside your plan.
A common Docker exercise is a broken artifact: a Dockerfile that rebuilds from scratch on every commit, or a container that exits the moment it starts, and the question of what is wrong. Backend loops raise Docker inside a deployment discussion; DevOps, SRE and platform loops give it more room. The opening questions check whether you see a container as an ordinary process with a fenced-off view of the host, for example why localhost inside it does not reach a service on the host.
A junior answer says what a command does. A senior answer traces a symptom to its mechanism: why docker stop waits out the grace period before killing the process (a shell-form CMD makes the shell PID 1, and it does not pass SIGTERM on), why exit code 137 often means the kernel's OOM killer, why one early cache miss rebuilds every layer after it, and why mounting the Docker socket hands a container root on the host.
The plan builds before it runs. The image comes first because every later module operates on one: the Dockerfile, the layer model that decides what a rebuild costs, and the digests that let a registry serve exactly the bytes you pushed. Running Containers then treats the image as a process: how it starts, takes signals and hits its limits in the lifecycle, and how it gets reachable ports and state that outlives it.
Local Development combines those pieces: the inner development loop inside containers, and Compose, which declares builds, networks and volumes in one file. From there the plan turns to risk, with hardening asking what a default container may do to its host, and then to operations: the engine with its logs and disk, and triage that turns a symptom into the next command.
Namespaces, cgroups and the runtime chain come late on purpose: by then you have met an OOM kill, a PID 1 that ignores signals and a permission error on a mount, and the internals explain all three. The plan ends off the single host, with production deployment without an orchestrator and Swarm, the orchestrator built into the engine.
If Docker is one line in a backend job description, narrow level to the range you are applying at, raise importance to must know and study by frequency; the questions asked most sit in building images and running containers. For a DevOps, SRE or platform role, study by topics and go all the way through, because the senior questions live in hardening, operations and the internals. If you have only used Docker Desktop, learn [the Linux VM behind it](/topics/cloud-docker-engine-desktop-vm) before someone asks where your containers actually run. Swarm is the only scheduler the plan covers. Pods, deployments and the controllers that hold a Kubernetes cluster at its desired state belong to the [Kubernetes roadmap](/roadmaps/kubernetes).
If Docker is one line in a backend job description, narrow level to the range you are applying at, raise importance to must know and study by frequency; the questions asked most sit in building images and running containers. For a DevOps, SRE or platform role, study by topics and go all the way through, because the senior questions live in hardening, operations and the internals. If you have only used Docker Desktop, learn the Linux VM behind it before someone asks where your containers actually run.
Swarm is the only scheduler the plan covers. Pods, deployments and the controllers that hold a Kubernetes cluster at its desired state belong to the Kubernetes roadmap.