Repository navigation
Feature request: Bucket lifecycle management via CRD #249
Description
Summary
Add bucket lifecycle management (ILM) to the RustFS operator via a dedicated CRD or lifecycle field in spec.buckets.
Use Case
We need automatic lifecycle management for production deployments:
- Log rotation: delete objects older than 30 days
- Backup retention: keep daily backups for 7 days, weekly for 4 weeks
- Temporary artifacts: delete CI/CD artifacts after 14 days
- Compliance: enforce data retention policies
Current Situation
- RustFS core supports lifecycle via admin API/CLI
- RustFS operator does NOT support lifecycle via CRD
Current Workaround
Managing lifecycle via CronJob + CLI script:
- Breaks GitOps workflow
- Requires additional maintenance
- Makes configuration drift likely
Request
Add lifecycle management via one of these approaches:
Option A: Dedicated CRD
Request
Add lifecycle field to spec.buckets in Tenant CRD:
apiVersion: rustfs.com/v1alpha1
kind: Tenant
metadata:
name: production
spec:
buckets:
- name: logs-bucket
objectLock: false
lifecycle:
rules:
- id: expire-old-logs
prefix: ""
status: Enabled
expiration:
days: 30
- id: cleanup-incomplete-uploads
status: Enabled
abortIncompleteMultipartUpload:
daysAfterInitiation: 1Metadata
Metadata
Assignees
Labels
No labels
Activity
GatewayJ commentedon Sep 12, 2026
Implemented and merged in #250.
The PR also updates the English and Chinese operator user guides, the Helm deployment README,
examples/provisioning-tenant.yaml, and the changelog.After installing the updated Tenant CRD and an Operator build from the updated
mainbranch, lifecycle rules can be declared directly underspec.buckets[]:Current behavior:
state: Presentrequires at least one rule.Every rule requires a unique
id,EnabledorDisabledstatus, and an explicitfilter.prefix;prefix: ""selects all objects.A rule must define
expiration.days,abortIncompleteMultipartUpload.daysAfterInitiation, or both. Day values must be positive.To explicitly remove an Operator-owned lifecycle configuration, use:
Omitting
lifecycleperforms no Lifecycle API calls and leaves the live bucket configuration untouched.The Operator can adopt an identical live configuration. It reports a conflict instead of overwriting or deleting a different live configuration unless the stored ownership hash matches.
Reconciliation state and ownership hashes are exposed under
status.provisioning.buckets.The implementation and documentation are now available on the
mainbranch and will be included in a subsequent Operator release.