Summary
- Add startup probes to chart-owned workloads so slow initialization does not trigger premature liveness restarts.
- Use the backend process-only liveness endpoint for startup checks.
- Let backend, worker, and frontend startup probes succeed immediately when healthy while preserving their maximum boot budgets.
- Standardize Fusionfire startup and readiness behavior.
- Enable a startup probe for the bundled Redis instance and clarify that it is intended for development and evaluation.
Upgrade notes
No configuration changes are required.
When upgrading from 0.13.33 or earlier with bundled Redis enabled, Kubernetes replaces the single Redis pod because its probe configuration changed. Expect a brief Redis interruption. Production installations should disable bundled Redis with logfire-redis.enabled=false and configure redisDsn for managed Redis.
PR: #195
Helm chart for self-hosted Pydantic Logfire