Sobes.tech
Interview Help
Resources
Interview Preparation
Resources
Job Search
Resources
Legal
Back to questions
Middle

How is caching implemented in Docker?

sobes.tech AI

Answer from AI

Caching in Docker is based on layers during image build. A Dockerfile consists of instructions, each creating a new layer. During subsequent builds, if an instruction and its context haven't changed, Docker uses the existing layer from the cache instead of executing the instruction again.

Factors affecting cache invalidation:

  • Instruction change: Any modification in the instruction itself (e.g., RUN apt-get update to RUN apt-get install).
  • Context change: Changes in files or directories used by the instruction (COPY, ADD). For example, changing the contents of a file being copied into the image.
  • Instruction order: Changing the order of instructions in the Dockerfile.
  • Using --no-cache: Explicitly disabling caching for the entire build.

Caching process:

  1. Docker reads the Dockerfile from top to bottom.
  2. For each instruction, it checks if there is an existing layer in the local cache that matches exactly the same previous instruction and context.
  3. If a match is found, Docker reuses that layer and moves to the next instruction.
  4. If no match is found (cache invalidated), Docker executes the instruction, creates a new layer, and adds it to the cache. All subsequent instructions after an invalidated layer will also not use the cache.

Example of instruction order impact:

Suppose you have a Dockerfile that first copies your code and then installs dependencies. If you often change the code but rarely change dependencies, caching the dependency installation will be useless because each time the code changes, the cache will be invalidated at the COPY step.

# Dockerfile 1 (Poor caching with frequent code changes)
COPY . /app
RUN pip install -r requirements.txt

Changing the instruction order can improve caching:

# Dockerfile 2 (Good caching with frequent code changes)
COPY requirements.txt /app/
RUN pip install -r /app/requirements.txt
COPY . /app

In this case, if only the code changes (not requirements.txt), the cache for RUN pip install will be preserved.

ADD and COPY instructions use content hash of files for cache invalidation.

The docker build command uses cache by default. To disable caching, the --no-cache flag is used.

# Build using cache (default)
docker build -t myimage .

# Build without cache
docker build --no-cache -t myimage .
PreviousNext