Closed lucky-lusa closed 3 years ago
Is Redis acceptable here as well? The data store basically just stores binary blobs, using a SQL database would increase the complexity as we'd need to define a schema and ensure it's up-to-date using migrations. Redis seems better suited to this since it's schema-less and well-suited for storage of binary data.
We can also use the file system if you tell me what the requirements are to be able to run multiple instances in parallel with the same file.
No worries, I have now implemented a Redis server as an alternative to the file store, we can still implement other data stores (the interface is very simple so it's quick to do). We can run multiple instances on the same file but only if the underlying file system supports atomic appends, which is not the case for most network file systems though.
We should change the file databases used so far to real SQL databases (preferably Postgres or a general implementation). This way we can ensure that the services scale horizontally and can be operated in parallel.