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

Dockerfile does not understand that it is necessary to clear the cache after executing the yum update command. Docker caching is based on the contents of commands in the Dockerfile.

When executing the RUN yum update command, Docker caches the result of this command. If the content of the RUN yum update line does not change, Docker uses the cached layer.

To ensure that the yum update command is executed during image build, you need to add something to this command that changes with each build or when an update is needed. One common way is to add a timestamp or a random number, although this is not always the best approach. A more proper method is to use multi-stage builds or manage dependencies more explicitly with a package manager.

However, the most effective way to forcibly clear the cache for a specific instruction is to change its content. Docker treats each instruction as an atomic unit for caching.

An example of such an "anti-cache" technique (not recommended for production but illustrates the principle):

RUN yum update -y && echo $(date +%s) > /tmp/update_timestamp

This method forces Docker to re-execute the layer on each build because the timestamp in the /tmp/update_timestamp file will change, altering the output of the RUN instruction and consequently the layer hash.

More proper approaches to ensure package freshness include:

• Multi-stage builds: Updating dependencies in an intermediate stage, then copying only the necessary artifacts into the final image.
• Explicitly specifying package versions: Fixing package versions in the Dockerfile or in a dependency list file and updating this file as needed.

The process when Docker clears the cache for an instruction works as follows:

1. Docker hashes the instruction in the Dockerfile (including arguments).
2. Docker searches for a layer in the local cache with the corresponding hash.
3. If the layer is found, it is used (cache hit).
4. If the layer is not found, the instruction is executed, a new layer is created, and this layer is added to the cache (cache miss).
5. On a cache miss, Docker also clears the cache for all subsequent instructions in the Dockerfile, ensuring their execution.

Thus, yum update by itself does not trigger a cache clear. This only happens when the RUN yum update instruction itself changes or when the cache is cleared by a previous instruction.