Data Model orchestration

[beasteers]

I'm trying to deploy directus as part of an automatic CI/CD pipeline (with terraform) so that instances can be brought up and down automatically.

Ideally, I'd like to synchronize directus's snapshot so that it can be instantiated from a github repo. I'm thinking of adding some sidecars such as git-sync and another sidecar that runs npx directus schema snapshot --yes ./snapshot.yaml periodically to dump the current schema so that it can be picked up by git-sync and uploaded to github.

So all that needs to be done to the chart for that part is to add the ability to add sidecars to the deployment.

---

But the second part is how to apply the snapshot to directus. The obvious answer is to use an initcontainer and to mount the snapshot as a config map, but it would need to be careful to not overwrite changes that a user makes through the UI.

That leads me to think that there is room for an extension that synchronizes the schema with a local file

• Listen to data model change events -> write snapshot to file
  • (I don't see data model changes in the events listed 😥)
• Watch for file change events -> apply snapshot
  • (Can use kiwigrid/k8s-sidecar with REQ_URL and the extension could have a custom endpoint to do the apply)

Then you'd use git-sync to sync the file with github.

I'm wondering if anyone else has a similar use case?

tl;dr - looking for a robust way to sync the directus data model with git

[kyuumeitai]

I'm a total k8s noob but I was wondering the same thing.

[kyuumeitai]

Hi @beasteers ! is this relevant to what you're trying to achieve?

directus/directus#14713

[beasteers]

Oooo that's great! I love when features you've been searching for were merged "last week" 🌈 Thanks so much for the heads up 🙏🏻

There's still a question of synchronization, but the API endpoints make it much easier to decouple that functionality from the container.

[fr3fou]

Have you figured anything new out yet? @beasteers
I was thinking of a similar idea, but instead of using sidecar for applying the schema snapshots, I'd use ArgoCD's PreSync resource hook https://argo-cd.readthedocs.io/en/stable/user-guide/resource_hooks to run a container which applies the snapshot using the Directus API.

I haven't figured out a way to reliably dump the snapshots though -- the schema can be changed both in a remote deployment (e.g staging / dev) or locally.