This vault has two jobs: teach the mechanisms behind devata, and preserve enough context to rebuild the platform later. Those jobs need different kinds of notes. A dated migration record should not silently become a runbook, and a concept note should not depend on a cluster still having the state that first exposed the concept.

Pick a path

If you want to…Start withFinish when you can…
Understand the whole systemdevata-system-mapdraw the control and data boundaries from memory
Rebuild the clusterreconstructing-devataname every authoritative input and off-cluster dependency in order
Learn Kubernetes through the real labkubernetes, then concept-mapexplain the controller loop and trace one request and one persistent write
Study operations and failure reasoningrecovering-loki-wal-replay-oom, then deleting-the-hypervisorseparate symptom, current failure loop, original trigger, and final proof
Follow the chronologyjourneyconnect dated decisions to their current successor design
Find curriculum gapskubernetes-coveragechoose the next gap that can be grounded in real devata work

Read, observe, explain, reconstruct

Use four passes instead of reading straight through:

  1. Read: state the problem and predict the mechanism before following commands.
  2. Observe: run only the commands whose context and safety boundary you understand.
  3. Explain: close the page and draw or say the causal chain in your own words.
  4. Reconstruct: name the declared input, controller, runtime object, proof, and rollback.
flowchart LR
  Read --> Predict
  Predict --> Observe
  Observe --> Explain
  Explain --> Reconstruct
  Reconstruct -. gap found .-> Read

Know which kind of page you are reading

  • Architecture explains the current durable shape. It should change when the design changes.
  • Concept explains a portable mechanism and uses devata as evidence.
  • Lab is safe to repeat and states context, mutation risk, expected output, and cleanup.
  • Case study is a dated record. Commands show what happened, not automatic permission to repeat it.
  • Operations gives an ordered recovery or reconstruction procedure with gates.
  • Planning records chronology and possible work. It is never the source of truth for the running cluster.

Commands are evidence until their safety box says otherwise

Always confirm kubectl config current-context, the repository, and the target node before using a command from a historical page. A command that was correct during a migration can be destructive against the completed system.

The source hierarchy

When pages disagree, use this order:

  1. live read-only observation for what is running now;
  2. the current default branch of the lab repository for intended state;
  3. architecture and operations pages for the durable model;
  4. dated case studies and the journey for historical reasoning;
  5. planning notes for possibilities, not facts.

The next page is devata-system-map.