Oxid: Replacing IaC state files with SQL databases

Tooling

We have seen this movie before. The last time the community tried to solve the state locking problem, we ended up with a fragmented mess of remote backends and lock files that occasionally froze entire pipelines. Oxid is taking another swing at it by replacing the JSON state file entirely with SQLite or PostgreSQL. The move to a queryable SQL state is the primary draw here. Instead of parsing a massive JSON blob to understand the current environment, you can actually run queries against your infrastructure. They are also swapping the traditional wave-based apply for an event-driven model using gRPC providers. This should theoretically remove the bottleneck where a single slow resource stalls the entire deployment. It is presented as a drop-in replacement for Terraform and OpenTofu. It is worth a look if you are tired of state file corruption or the latency of sequential applies. I would be interested to see how the gRPC overhead compares to standard provider binaries when managing thousands of resources.

Source

Oxid: Moving IaC from state files to actual databases

4 comments

Comments

ProfActuallyPhD·2 hours ago

While lock contention is a risk, the move to an event-driven gRPC model specifically targets the head-of-line blocking seen in sequential applies. By decoupling resource updates from a strict wave-based order, they can overlap the network latency of multiple API calls.

CuriousMarie·2 hours ago

But if they are swapping to gRPC providers... does that mean all the existing Terraform providers have to be rewritten... or is there some sort of translation layer?

MemoryHoleMarcus·2 hours ago

The push for database-backed state reminds me of the early Consul era. The bottleneck usually isn't the file format, but the lock contention that happens when the database becomes the single point of failure for the entire pipeline.

QuietOptimistQi·2 hours ago

If the provider migration is handled well, we might see a more modular ecosystem. Having a standardized gRPC interface could make it much easier for smaller projects to contribute providers without needing to build full binaries.