Start here
Choose your path
Trellara is a verification and operations layer above PostgreSQL logical replication. It is not a replacement for it. These docs are organised around three questions, because in practice everyone arrives with one of them.
Is this safe to point at my database?
You want to know what it touches, what it installs, and what the first command can do before you run anything. About twelve minutes, and it writes nothing.
trellara-check $POSTGRES_URL DeveloperWhat may I assume about the stream?
You are building a flow or a consumer and need to know exactly what is guaranteed, what is configurable, and where the protocol fails closed.
docs/DESIGN.md OperatorWhat do I need before this runs on call?
The production configuration gate, the Kafka contract, secrets, the health endpoints and the metric names — then quarantine and reseed for when it breaks.
trellara status --view diagnosticsWhat exists today
Being direct about this up front saves you the click. Trellara is pre-1.0.
Start here and Concepts are complete and current.
Operations has three of its eight pages — production
deployment, security and secrets, and monitoring and alerts — because those
three describe contracts the code enforces today. The other five need operating
experience that does not exist yet, and writing them from imagination would be
worse than leaving them out. Build on Trellara has the protocol
specification; the conformance suite it should sit beside has not been built.
Reference is generated from the repository by
tools/gen_reference.py — the CLI definitions, the
config structs, the metric constants and the protobuf schema — so those six
pages cannot drift from the code without the generator changing too. The error
catalog is the one entry still missing, because it has no single source in the
code to generate from yet. Everything unwritten is dimmed in the sidebar rather
than stubbed with a placeholder page.
The honest inventory of what has and has not been established lives on the evidence table. Read it before you plan around anything here.
How the docs are shaped
| Section | Form | Assumes |
|---|---|---|
| Start here | Tutorial — one managed path | Nothing. The author owns your success. |
| Concepts | Explanation — why it works this way | You have read a replication doc before. |
| Operations | How-to — branching, task-shaped | Competence. You know what you are trying to do. |
| Build on Trellara | Reference + explanation | You are writing code against the protocol. |
| Reference | Generated from the code, exhaustive, dry | You are looking up one specific thing. |
| Trust and evidence | Artifacts, not prose | You are being asked to justify a choice. |
A tutorial and a how-to are never mixed. A tutorial has one path and no branches; a how-to assumes you can handle branches and shows you all of them. Mixing the two produces a document that fails both readers.
A note on one word
These docs never say “the Trellara operator.” That phrase collides with the Kubernetes sense of the word, and Trellara does not ship one. Operator here always means the person on call.