bc-gov-emerald

name

bc-gov-emerald

description

BC Government Emerald OpenShift platform mechanics: namespace conventions, AVI InfraSettings route annotations, DataClass pod labels, Helm openshift mode, StorageClass, and DNS split-tunneling. Use when creating or reviewing Helm charts, Routes, or deployment manifests targeting Emerald specifically. For zone/classification decisions see bc-gov-sdn-zones; for NetworkPolicy YAML patterns see bc-gov-networkpolicy.

BC Gov Emerald Platform Standards

Emerald-specific mechanics only. Platform-independent concepts live in companion skills:

• Zone model, ISCF classification, internet egress constraints → bc-gov-sdn-zones
• NetworkPolicy YAML patterns and two-policy rule → bc-gov-networkpolicy
• End-to-end network architecture reasoning → bc-gov-network-architect agent

---

Namespace Convention

<license>-dev     ← development
<license>-test    ← staging / QA
<license>-prod    ← production
<license>-tools   ← CI/CD tooling, Artifactory, Vault

---

AVI InfraSettings — Route Annotation

Controls which VIP pool handles the Route. Get this wrong and traffic silently drops.

Annotation value	VIP	When to use
dataclass-medium	Private VIP — VPN only	✅ All internal workloads (default)
dataclass-high	Private VIP — sensitive data	Higher-trust internal workloads
dataclass-public	Public internet VIP	Internet-facing routes with public exposure
dataclass-low	⚠️ NO VIP on Emerald	NEVER USE — DNS resolves but ERR_EMPTY_RESPONSE

# Required on every OpenShift Route
metadata:
  annotations:
    aviinfrasetting.ako.vmware.com/name: "dataclass-medium"

AKO re-adds this annotation within ~15 seconds if removed — always keep it in Helm values.

---

DataClass Pod Label

Pod DataClass label must match the AVI annotation suffix.

podLabels:
  DataClass: "Medium"   # matches "dataclass-medium" annotation

Mismatch rule: DataClass: Low + dataclass-medium route → SDN silently drops traffic.

Required pod labels (enforced by ag-devops Datree + Conftest)

Every Deployment/StatefulSet pod template must carry all three labels. Missing any will cause the ag-devops policy gate to deny the manifest.

podLabels:
  DataClass: "Medium"          # Low | Medium | High
  owner: "<team-or-ticket>"    # team name or Jira ticket reference
  environment: "development"   # production | test | development (exact values)

When using ag-template.deployment, set ModuleValues.dataClass (renders DataClass) and add owner/environment via a LabelData fragment:

{{- define "myapp.labels" -}}
owner: jag-pssg-team
environment: development
{{- end }}
# ... in dict: set $p "LabelData" "myapp.labels"

Internet-Ingress label

podLabels:
  Internet-Ingress: "DENY"    # default — correct for all internal services
  # Internet-Ingress: "ALLOW" # only if reachable from public internet via Public VIP

For the full ISCF → DataClass mapping and zone egress constraints, see bc-gov-sdn-zones.

---

NetworkPolicy Model

Emerald default-denies both Ingress AND Egress. See bc-gov-networkpolicy for full YAML patterns, the two-policy rule, DNS egress, CIDR egress, and the debugging checklist.

Quick reminder of the flows that need policies on Emerald:

Flow	Policy needed
Router → Frontend	Ingress on Frontend
Router → API	Ingress on API
Frontend → API	Ingress on API + Egress from Frontend
API → DB	Ingress on DB + Egress from API
Any pod → DNS	Egress UDP+TCP 53 on every pod

---

Route Edge Termination (Conftest hard-deny)

ag-devops Conftest denies edge-terminated Routes unless the Route has either:

• Label app.kubernetes.io/component: frontend, or
• Annotation isb.gov.bc.ca/edge-termination-approval: "<ticket>"

Passthrough (spec.tls.termination: passthrough) and re-encrypt termination are not affected.

# If edge termination is approved:
metadata:
  labels:
    app.kubernetes.io/component: frontend   # OR:
  annotations:
    isb.gov.bc.ca/edge-termination-approval: "ISB-12345"

---

PriorityClass (Polaris priorityClassNotSet check)

Every Deployment and StatefulSet must reference a PriorityClass. Polaris issues a failure if spec.template.spec.priorityClassName is unset. Define one PriorityClass per application group in the chart:

apiVersion: scheduling.k8s.io/v1
kind: PriorityClass
metadata:
  name: <app>-priority
value: 1000000
globalDefault: false

Then in each workload:

spec:
  template:
    spec:
      priorityClassName: <app>-priority

---

OpenShift Mode in Helm (global.openshift: true)

Set this in every Helm chart's values.yaml targeting Emerald:

global:
  openshift: true

Effect:

• Deployment/Job pod securityContext does not pin runAsUser/runAsGroup — OpenShift SCC assigns runtime UID/GID
• Adds checkov.io/skip999: CKV_K8S_40=... annotation to suppress Checkov false-positive
• Still enforces runAsNonRoot, allowPrivilegeEscalation: false, readOnlyRootFilesystem: true, capabilities drop ALL

⚠️ Omitting global.openshift: true when using the ag-helm library chart causes the deployment template to pin runAsUser: 10001 which may conflict with the namespace's SCC.

---

StorageClass

Choose based on access mode and workload type:

StorageClass	Access mode	Best for
netapp-file-standard	RWX (multi-pod)	Shared file storage, build artefacts, config mounts
netapp-block-standard	RWO (single-pod)	Databases, stateful workloads requiring higher IOPS

# For shared / file-based workloads:
storageClassName: netapp-file-standard

# For databases and stateful workloads (preferred for block I/O performance):
storageClassName: netapp-block-standard

netapp-block-standard is single-pod (RWO) — do not use for workloads that require concurrent access from multiple pods.

---

DNS Split-Tunneling

Route hostnames (*.apps.emerald.devops.gov.bc.ca) resolve only via BC Gov VPN DNS. Local / home DNS returns NXDOMAIN. Ensure VPN client routes this domain through VPN DNS before debugging route connectivity issues.