Does the backup actually come back?
Operational scar tissue turned into tools: the untested restore, the lost evidence, the diagram that no longer matches the cluster.
Before, during and after the incident
Most of these exist because something went wrong once and the tooling to understand it did not exist at the time. Anything not publicly released is marked.
Before
Prove the assumptions while it is cheap. A restore that has never run is a hypothesis, and a diagram nobody checks is fiction.
During
Collect the evidence before the fix destroys it. A support bundle taken in the first five minutes answers questions the post-mortem cannot reconstruct a week later.
After
The follow-up work that decides whether the same incident happens twice.
The papers
Every paper carries the methodology behind any number it emits, and what the tool deliberately does not do. Pick them up from the catalogue.
- Capturing the evidence before the fix destroys ita one-command Kubernetes support bundle
- Answering the operational questions engineers get wrong under pressureclient-side browser tools for platform engineers
- Architecture documentation that can be checked line by linea stdlib-only architecture mapper with a drift check
- Keeping the architecture diagram true to what is deployeda C4 container diagram generated from Terraform state and a live cluster
- Proving a rewrite behaves identically, on production traffica traffic-mirroring proxy with a structural response differ
- Turning Kubernetes usage data into a merged requests changea rightsizing report that emits committable patch YAML
- Proving the backup restores, nightly, with nobody watchinga scheduled restore-and-smoke-test harness
- Finding the configuration difference between two environments, safelyan environment-variable differ with secret masking on by default
- Keeping a working day's kubectl port-forwards alivea supervised kubectl port-forward manager
- The scheduled job that silently skips when the clocks changea cron translator with timezone-aware next runs and DST warnings
The questions in between, in your browser
Cache headers, CronJob schedules, probe timings, regex backtracking, NetworkPolicy reachability: one input, one clear answer, shareable by URL. No accounts, no backend, no telemetry.
Working on this?
Tell me what you are looking at and I will tell you honestly whether any of this helps. No pitch attached, and the papers are free either way.