[Infrastructure] Missing resource limits, image version pinning, PDB, single replicas

[Flegma]

Summary

Multiple production-readiness gaps in Kubernetes workload configuration.

Findings

• All deployments/statefulsets — no resources.requests or resources.limits.
• API, Web, Connector, TimescaleDB, MinIO — all use :latest tags. No version pinning.
• No PodDisruptionBudgets for any workload.
• API, Hasura, Web — all replicas: 1. Single point of failure.
• No ResourceQuotas or LimitRanges in namespace.

Impact

Unlimited resource consumption. Unpredictable deployments. No HA guarantees.

Suggested Fix

Add resource limits. Pin to semantic versions. Create PDBs. Increase replicas. Add namespace quotas.