ServiceRadar v1.4.67

A new CNPG cluster created by the chart now bootstraps its
database instead of failing the core migrations hook, and core
health events stop reporting an hourly "healthy to healthy"
change.

Whats New

1.4.67

• Helm: a new CNPG cluster created by the chart now bootstraps
  its database. The cluster's initdb SQL created timescaledb,
  postgis and vector before the platform schema existed, so
  they landed in public, where the schema baseline cannot move
  them, and the serviceradar-core-migrations hook failed on
  platform.create_hypertable. The chart now runs no initdb SQL
  by default and leaves the core migrations job to create those
  extensions in platform. cnpg.extensionsSQL still renders
  when set, and an explicit [] now means none instead of
  falling back to the old list. Existing clusters and SPIRE-mode
  Postgres are unaffected (PR #461, issue #454).
• Core: health tracking no longer announces an unchanged state as
  a transition. The seasonal baseline producer records a heartbeat
  on every hourly run, and each one was published as "changed from
  healthy to healthy", which became a health.core.state_change
  event every hour. Unchanged records still land on the health
  timeline, so the baseline freshness check keeps working, but
  only a real change or a first record now publishes the health
  log, the event and the live broadcast. The heartbeat also
  reports the state it last recorded instead of always "healthy",
  so a repeated failure publishes nothing, a lasting failure keeps
  one core_health_check_unhealthy incident open instead of
  resolving and reopening it every hour, and a recovery still
  clears it. Events already recorded stay until retention removes
  them (PR #457).

Upgrade Notes

• If your values set cnpg.extensionsSQL, CNPG still runs that
  SQL when it creates a new cluster, and an override that copies
  the old unqualified CREATE EXTENSION list still leaves a new
  cluster unable to bootstrap. Drop the override, or run
  CREATE SCHEMA IF NOT EXISTS platform; first and create each
  extension WITH SCHEMA platform (age in ag_catalog).
  Existing clusters never re-run this SQL and need no action.