BizFirst Observe
Loki Log Retention
Loki manages log retention through its Compactor component. The compactor runs periodic cleanup jobs that delete log chunks older than the configured retention period. Retention can be set globally or per-tenant.
Global Retention Configuration
# loki-config.yaml — retention configuration
compactor:
working_directory: /loki/compactor
shared_store: s3
compaction_interval: 10m # How often compactor runs
retention_enabled: true # MUST be true to enable deletion
retention_delete_delay: 2h # Wait 2h after marking before deleting
retention_delete_worker_count: 150
# Global limits (applies to all tenants unless overridden):
limits_config:
retention_period: 720h # 30 days (720 hours)
# Alternative: 2160h = 90 days, 8760h = 1 yearPer-Tenant Retention
BizFirstGO is multi-tenant. You may need different retention periods per tenant — for example, financial tenants requiring 1-year log retention while development tenants need only 7 days:
# loki-config.yaml — per-tenant retention overrides
limits_config:
retention_period: 720h # Global default: 30 days
# Per-tenant overrides in a separate runtime config:
# loki-runtime-config.yaml
overrides:
# Financial tenant — 1-year log retention
"tenant-financial-corp":
retention_period: 8760h # 365 days
# Development tenant — 7-day retention to save space
"tenant-dev":
retention_period: 168h # 7 days
# Standard SaaS tenant — 90-day retention
"tenant-saas-standard":
retention_period: 2160h # 90 daysHow Loki Compaction Works
| Phase | What Happens | Timing |
|---|---|---|
| Compaction | Merges small chunks into larger ones for efficient query | Every 10 minutes (configurable) |
| Retention marking | Compactor marks chunks older than retention_period for deletion | During compaction run |
| Deletion delay | Marked chunks wait in "to-delete" state (allows abort if misconfigured) | 2 hours (retention_delete_delay) |
| Actual deletion | Chunks are permanently deleted from object storage | After delay expires |
Monitoring Retention
# Check Loki compactor status via metrics:
curl -s http://localhost:3100/metrics | grep loki_compactor
# Key metrics:
# loki_compactor_runs_total — total compaction runs
# loki_compactor_blocks_cleaned_total — chunks deleted by retention
# loki_compactor_retention_marked_for_deletion_total — chunks marked
# Verify retention is working (query should return no data for period before retention):
# In Grafana Explore (Loki):
# Set time range to older than your retention period
# Run any LogQL query
# Expected: "No data" (logs have been deleted)
Retention Is Irreversible
Once logs are deleted by the Loki compactor, they cannot be recovered. Ensure your retention period covers all compliance requirements before enabling deletion. For compliance-sensitive deployments, archive to cold storage (S3 Glacier) before the hot retention period expires — see the Archiving page for configuration.