Crash-loop protection and deployment fixes
Services stuck restarting can be paused automatically, alongside fixes for runtime metrics, server binding and reused build images.
Stop sustained crash loops
A service that repeatedly restarts with no healthy replicas can now be automatically paused. This protection remains active even when health alerts are disabled. Services with a healthy replica are not paused by this safeguard.
Inspect the failed startup and runtime logs, correct the application configuration, then redeploy or resume the service. The monitor preserves the crash-loop state across restart observation windows instead of resetting protection too early.
More reliable operation
- Runtime CPU and memory usage is compared with the container’s configured limits.
- Server binding defaults were corrected so application processes can be reached by the platform.
- Reused build artifacts are no longer deployed with a legacy registry hostname.
- Registry cleanup preserves images that running workloads still reference.
- Customer Bun builds default to Bun 1.4.2 unless the repository pins a version.
See the deployment lifecycle documentation for where to inspect build output, health and runtime logs.