Field notes

Notes on Kubernetes fleets, remediation pipelines and infrastructure contracts

Written from production platform work, not summaries of documentation. Each note comes out of a decision that had a cost attached to it.

• June 20266 min read

  Detection first: remediating Kubernetes upgrades without mutating clusters

  Why post-upgrade Kubernetes remediation across a large fleet should start with a read-only detection stage that never writes, and how that boundary changes the blast radius of automation.

  KubernetesPlatform EngineeringAutomation
• July 20265 min read

  JSON contracts between pipeline stages: treating remediation as data

  Positional shell output breaks silently. Typed JSON records at every stage boundary make an infrastructure remediation pipeline reviewable, diffable and independently versionable.

  AutomationInterfacesReliability
• July 20265 min read

  Exit code zero is not a success signal: closing the remediation loop

  Infrastructure automation reports success far too easily. Re-running detection with the same classifier that raised the issue is the only honest way to prove a fleet-wide fix landed.

  ReliabilityKubernetesVerification