Repository navigation
Resource Scheduler 0.8.16
Resource Scheduler 0.8.16
This release keeps the scheduler process alive during a planned control-plane restart without replacing its container.
Changes
- Reconnect the control-plane event subscriber after a clean disconnect or an error.
- Restart the metadata watcher in place after an error or a recovered panic.
- Keep
/healthcheckas process liveness and expose dependency status through/readiness. - Add regression coverage for reconnection, panic recovery, and liveness/readiness separation.
Verification
- Go 1.26.5 unit tests and
go vetpassed. - Linux race tests passed for all five packages.
- The package-script, build-policy, migration-policy, formatting, and history gates passed.
- A control-plane outage smoke test kept the same container ID running with restart count
0, returned HTTP200from/healthcheck, and returned HTTP503from/readinesswhile dependencies were unavailable. - Trivy 0.73.0 reported zero Critical or High image vulnerabilities and no detected image secrets. The attached JSON reports and CycloneDX SBOM preserve the evidence.
Artifacts
- Container image:
ghcr.io/pasturestack/resource-scheduler:v0.8.16 - OCI digest:
sha256:7d7d39c43d4290161001e09220ef5e37abe2837f9109929ee87ce4ade47e4353 - Source commit:
c54bc3afd9b83577be73049b88e63430488566b4
Use the numeric version tag in deployment configuration. Do not place the digest in the user-facing image field.