How does Dockerfile understand that it needs to clear the cache after executing 'yum update'?

Docker caches image layers based on the content of commands in the Dockerfile and their context. When executing the yum update command, Docker looks at the command line itself and the files that influence its execution.

If the yum update command changes the system state (for example, updates packages), but the command itself in the Dockerfile does not change, Docker by default uses the cache of the previous layer because it does not see any changes in the instruction.

To invalidate the cache after yum update, the usual approaches are:

• Combine commands that change the state into a single RUN to make caching more controllable.
• Add cache-cleaning commands for the package manager (e.g., yum clean all) to reduce the layer size.
• If you need to force cache invalidation, you can modify the command (for example, add an argument with the current date or a variable) so Docker considers it new and rebuilds the layer.

Example:

RUN yum update -y && yum clean all

To force cache invalidation:

ARG CACHE_DATE=2024-06-01
RUN yum update -y && yum clean all

Changing the value of CACHE_DATE will cause Docker to rebuild this layer, as the command now differs.

Thus, Docker recognizes the need to invalidate the cache based on changes in the command text and context, not on internal system state changes.