Discussion: Improvement suggestions

[A7exSchin]

Hi there RagRug team,

we are currently aiming at using the RagRug toolkit for situated data presentation. While trying to set up the environment we encountered several issues:

1. The current backend setup seemed to only work on unix hosts with a certain user setup. The dependency on the host filesystem for mounts and permissions is not considering Docker best practices. Windows users using MingW64 are out of luck, as $(pwd) is resolves the mount path to ../mnt;C. I see that is no particular issue with ragrug, but docker offers tools for platform independency (see later)
2. The RagRug container includes modules that require a higher version (v12) than it installs (v10).
3. The RagRug setup Script references an old file rr_002.json instead of rr_last.json
4. The used node-red-contrib-cloudantplus module is deprecated, replaced with node-red-node-cf-cloudant
5. Using docker-compose makes a lot of the scripting unnecessary. Compose offers tooling to set up a system of containers as well as shut them all down.

We have forked the repo and sketched the docker-compose structure as well as updated the NodeJs version and the cloudant module (as well as the flows in the rr json). The changes also make RagRug executable on windows machines with git bash.

The updated fork can be found here: https://github.com/A7exSchin/ragrug

If there is anything more that comes to our attention I will update the issue.

P.S: Due to the compose structure some of the hostnames are not updated, yet

Edit 1:
Another questions popped up.
In the build-and-run-local-docker-dev.sh file you are mounting the whole root/backend/mnt folder into the ragrug container. Why is this necessary?

[philfleck]

Hi Alexander,
thank you very much for your feedback. And also thank you for forking and participating.

The documentation is still lacking information. The intention for windows was to run it using WSL2 (eg using ubuntu) and Docker for Windows and therefore the scripts would always run under windows.

at 2) this came recently to our attention, which is will be fixed.

at 3) thanks for letting us know, we will address that. We will likely remove the version and only keep the last one, so it becomes clear. This problem has some historic issues.

at 4) this also came recently to our attention, thank you a lot for noticing.

at 5) We known about that fact, and are glad that you took a run on it. The main reason is, that in our environment it runs the parts on multiple machines which was easy for us go with the scripts. I guess this can also easily be using docker compose and to turn of some components.

However, we really appreciate your efforts.

at Edit 1:
this has some reasons:

• some tasks working on larger files require a shared space
• there is more work in progress in our pipeline which is not published which requires a shared storage
• bare-metal applications might need to push files into the docker environment
• it keeps it simple with backup (eg couchdb, postgres, node-red) allows to zip the mnt folder

but also: not all containers need access to all, therefore this could be narrowed down to be more restrictive.

I hope this helps and many thanks for the discussion.
please let me know.

best,
pf