Handle the case where customized configuration exists prior container start

[petr-stratoscale]

The container's entry point currently always overwrites the configuration file.
This does not work well in the scenario where a customized config file is stored on a mounted volume. Any customization are then lost when the container starts.

The entry point should detect and handle (like for example Mysql and Postgres) this case.

[yosifkit]

How do you propose we detect that the config is not just the default config that we need to sed to the proper values?

On the other hand, since we don't have the same initdb.d directory like postgres and mysql the user can just skip the provided entrypoint when they supply their own config file.

[petr-stratoscale]

We could check the data directory and skip config if it has already been initialized (cassandra-created subdirs exist).

@yosifkit How can I skip the entry point?

Thanks.

[yosifkit]

To skip the entrypoint, basically just replace the entrypoint and command with the current command: CMD ["cassandra", "-f"]

$ docker run --entrypoint cassandra cassandra:3.11 -f

You'd still need to build or map in the config file.

[tianon]

Closing given that there's a workaround that's simpler (and more reliable) than trying to detect the default configuration file. 👍