Kubernetes Deployment Principles

This repository does not ship a full Helm chart in docs/v3, but the following principles are MonkDB-specific and production-safe.

Workload type

Use StatefulSet for MonkDB nodes.

• Stable identity per pod.
• Persistent volume per pod.
• Ordered rollout control.

Config and flags

You can configure MonkDB through:

• monkdb.yml mounted via ConfigMap.
• CLI flags (-Ckey=value) in container args.
• Environment variables for selected runtime values (MONKDB_HEAP_SIZE, etc.).

Example args pattern:

args:
  - -Ccluster.name=prod-cluster
  - -Cnode.name=$(POD_NAME)
  - -Cnetwork.host=0.0.0.0
  - -Cdiscovery.seed_hosts=monkdb-0.monkdb,monkdb-1.monkdb,monkdb-2.monkdb
  - -Ccluster.initial_master_nodes=monkdb-0,monkdb-1,monkdb-2

Storage

• Use SSD-backed PVCs for hot data.
• Separate blob/data paths if required by workload economics.
• Define storage classes with predictable latency.

Scaling rules

• Add nodes gradually.
• Watch shard relocation and query latency during scale events.
• Avoid frequent oscillation in replica/shard topology.

Security

• Use Kubernetes secrets for credentials and certificates.
• Restrict service exposure with NetworkPolicies/Ingress rules.
• Prefer mTLS/TLS end-to-end in production.