When split DNS made a Firebase site appear offline
Resolver comparison exposed an over-broad private DNS redirect while the public deployment remained healthy.
Read the incident analysis
Operational Writing
Incidents and engineering decisions, explained concisely.
These articles preserve the reasoning behind diagnosis, maintenance, networking, recovery, and architecture decisions without exposing private infrastructure.
Why kubectl drain is not enough
Drain success does not prove replacement placement, endpoint continuity, storage safety, quorum, or recovery.
Read the maintenance articleBackup completion does not prove recoverability
A completed backup job does not prove that data, identity, storage, and bootstrap dependencies restore in order.
Read the recovery articleTracing DNS to Pod in Kubernetes
Following each request-path handoff reveals where declared intent first diverges from observed behavior.
Read the networking articleHow to review the engineering evidence
A guided route through platform boundaries, service routing, recovery proof, and lifecycle decisions.
Open the evidence guide