Operations¶
Keeping a deployment healthy and knowing when it is not.
- Health and Readiness — the probes and what they mean
- Monitoring — metrics, logs, and what to alert on
- Backup and Restore — taking backups that work
- Integrity Verification — proving the bytes are intact
- Capacity Planning — logical, physical, and what grows
- Disaster Recovery — when things go badly wrong
Daily¶
record-store status --endpoint https://management.example.com
record-store storage inspect --endpoint https://management.example.com
If those two look right, the deployment is fine. If they do not, the rest of this section says what to do.
Signals worth watching¶
| Signal | Where | Concerning when |
|---|---|---|
| Readiness | /ready |
Anything but 200 |
| Disk headroom | A host exporter on the data directory's filesystem | Below 20% |
| Error rate | record_store_errors_total / record_store_requests_total |
Above a few percent |
| Missing payloads | metadata_without_data in storage inspect |
Above 0 |
| Storage growth | record_store_storage_physical_bytes |
Outpacing the disk you have |
| Audit denials | record-store audit --limit 100 |
A burst from one principal |
Nothing is pruned for you¶
Two stores grow without bound and have no automatic retention:
- The audit trail grows with request volume.
- Version history grows with every overwrite on a versioned bucket.
Lifecycle rules handle the second. The first is a capacity-planning input. See Capacity Planning.