Monorepo vs Multi-Repo Architecture - Seeking Community Insights on CDF Toolkit Management

Question

Hi Everyone,We're currently managing multiple CDF Toolkit solutions in a mono-repo and evaluating whether to maintain this approach or migrate to a multi-repo structure. We'd love to learn from teams who've navigated through this decision.Current Setup (Mono-repo):1. Single repository housing multiple solutions with shared infrastructure. 2. Shared auth-governance across all solutions.3. Teams need independence but also need to prevent cross-team overwrites.We're Looking For: 1. Real-world examples: How did you decide between mono/multi-repo? What factors mattered most?2. Tooling patterns: How do you handle shared resources (auth governance) in multi-repo? How do you manage CDF Toolkit versioning across repos?3. At how many solutions/teams did you reach the limit of mono-repo and decide to switch to multi-repo?4. How do teams with multi-repos handle toolkit version synchronization? Do you enforce a minimum/maximum version across repos?5. Documentation (runbooks or migration guidelines)We'd love feedback from: ​@Markus Pettersen CC: ​@Marius Sørenes , ​@Peter Arwanitis  Any experiences or documentation would be valuable.Thank you,David Shaji George

Markus Pettersen · Accepted Answer

As mentioned, we currently use a multi-repo structure but have decided to move to a mono-repo approach as part of our transition from ADO to GitHub for our CDF Toolkit repositories.

We are still in the early stages of implementation, so it’s too soon to share concrete experiences with the mono-repo setup.

One of the main challenges with our multi-repo setup has been governance, particularly around access management and admin settings in CDF (such as location filters). This is something we plan to centralize regardless, as allowing access configuration from any repo poses a security risk.

We are also moving away from CDF classic datasets toward data modeling. Our current design, which assumes one repo per dataset, is already outdated and needs to be revised. With data modeling and the introduction of data products/domains, our approach to using the Toolkit will need to evolve anyway. For now, we’ve chosen a mono-repo structure, but whether this is the right decision, and how it works once developers start using it, is still too early to tell.

Peter Arwanitis · Answer

hi ​@Markus Pettersen, I’ve mentioned to Equinor that Aker BP went through this investigation lately as you’re still using cdf-toolkit in a multi-repo setup, and maybe you can share some of the insights and decisions made with the community?

Sign up

Welcome to Cognite Hub