fix(kilobase): fix hourly backup bug and add WAL retention cleanup

[h0lybyte]

Summary

• Fixes backup running every hour instead of daily — CNPG uses 6-field cron (with seconds), so 0 2 * * * was parsed as sec=0 min=2 hour=* (hourly at :02). Fixed to 0 0 2 * * * (daily at 2:00 AM UTC).
• Adds WAL retention cleanup — instanceSidecarConfiguration.retentionPolicyIntervalSeconds: 1800 runs barman cleanup every 30 minutes, removing WAL files and base backups beyond the retention window.
• Increases retention from 3d to 7d — keeps a full week of backups for safety.

Root Cause of S3 Cost Increase

The 5-field cron 0 2 * * * was creating 24 base backups per day instead of 1. Combined with no WAL cleanup, S3 storage was growing unbounded. This has been running since the ScheduledBackup was created (~101 days ago).

Post-merge Manual Steps

After ArgoCD syncs, clean up the 29 failed backup CRDs:

kubectl get backups.postgresql.cnpg.io -n kilobase --no-headers | grep failed | awk '{print $1}' | xargs kubectl delete backups.postgresql.cnpg.io -n kilobase

The WAL cleanup sidecar will automatically start purging old WAL files from S3 within 30 minutes of deployment.

Test plan

• ScheduledBackup shows nextScheduleTime ~24h from now (not ~1h)
• No new backup created at the next hour mark
• WAL cleanup runs within 30 minutes (check barman-cloud sidecar logs)
• S3 bucket size starts decreasing over the next few hours

🤖 Generated with Claude Code

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None