chore(querybook): set pod security standards and adjust resources

querybook/README.md

@@ -350,6 +350,62 @@ Trino (HTTPS via external hostname)
 - **Trino Connection**: Must use external HTTPS hostname (not internal service name)
 - **User Impersonation**: Admin credentials with X-Trino-User header for query attribution
 
+## Pod Security Standards
+
+**Current Configuration**: `privileged` (enforce) / `baseline` (warn, audit)
+
+Querybook namespace is configured with the following Pod Security Standards:
+
+```yaml
+pod-security.kubernetes.io/enforce: privileged
+pod-security.kubernetes.io/warn: baseline
+pod-security.kubernetes.io/audit: baseline
+```
+
+### Why Not Restricted or Baseline?
+
+Querybook's embedded **Elasticsearch** component requires privileged containers and special Linux capabilities that violate both `restricted` and `baseline` Pod Security Standards:
+
+**Elasticsearch Requirements**:
+
+- `privileged: true` - Container must run in privileged mode
+- `capabilities.add: [IPC_LOCK, SYS_RESOURCE]` - Requires Linux capabilities for memory locking
+- `sysctl -w vm.max_map_count=262144` - Init container needs privileged mode to configure kernel parameters
+
+These requirements are necessary for Elasticsearch to:
+
+1. Lock memory to prevent swapping (performance)
+2. Set virtual memory map count (stability)
+3. Configure ulimit for unlimited locked memory
+
+**Security Implications**:
+
+- Elasticsearch containers run with elevated privileges
+- Init containers can modify kernel parameters
+- Other components (web, worker, scheduler, redis) run without special privileges
+
+**Mitigation**:
+
+- `warn` and `audit` at `baseline` level to track violations
+- Web init container (copy-keycloak-auth) uses `restricted`-level security context
+- Future: Consider external Elasticsearch service to enable stricter Pod Security Standards
+
+**Component Security Status**:
+
+| Component      | Privileges Required | Security Level |
+|----------------|---------------------|----------------|
+| Elasticsearch  | privileged=true, IPC_LOCK, SYS_RESOURCE | Violates baseline |
+| Web            | None (container), runAsNonRoot (initContainer) | Baseline-ready |
+| Worker         | None | Baseline-ready |
+| Scheduler      | None | Baseline-ready |
+| Redis          | None | Baseline-ready |
+
+To check current Pod Security Standards configuration:
+
+```bash
+kubectl get namespace querybook -o jsonpath='{.metadata.labels}' | jq
+```
+
 ## Authentication
 
 ### User Login (OAuth2)

querybook/justfile

@@ -16,10 +16,20 @@ export KEYCLOAK_HOST := env("KEYCLOAK_HOST", "")
 default:
     @just --list --unsorted --list-submodules
 
-# Create Querybook namespace
+# Create Querybook namespace with Pod Security Standards
+# Note: Elasticsearch requires privileged containers, so enforce=privileged
+# but warn/audit at baseline level to encourage security improvements
 create-namespace:
-    @kubectl get namespace ${QUERYBOOK_NAMESPACE} &>/dev/null || \
+    #!/bin/bash
+    set -euo pipefail
+    if ! kubectl get namespace ${QUERYBOOK_NAMESPACE} &>/dev/null; then
         kubectl create namespace ${QUERYBOOK_NAMESPACE}
+    fi
+    kubectl label namespace ${QUERYBOOK_NAMESPACE} \
+        pod-security.kubernetes.io/enforce=privileged \
+        pod-security.kubernetes.io/warn=baseline \
+        pod-security.kubernetes.io/audit=baseline \
+        --overwrite
 
 # Delete Querybook namespace
 delete-namespace:
@@ -254,22 +264,6 @@ install:
     kubectl wait --for=condition=Available deployment/web \
         -n ${QUERYBOOK_NAMESPACE} --timeout=300s
 
-    echo "Waiting for service to be accessible at https://${QUERYBOOK_HOST} ..."
-    for i in {1..60}; do
-        http_code=$(curl -k -s -o /dev/null -w "%{http_code}" https://${QUERYBOOK_HOST} 2>/dev/null || echo "000")
-        if echo "${http_code}" | grep -q -E "200|302|401|403"; then
-            echo "Service is now accessible (HTTP ${http_code})"
-            break
-        fi
-        if [ $i -eq 60 ]; then
-            echo "Warning: Service may not be fully accessible yet (last status: ${http_code})"
-            echo "Please wait a few more minutes and try accessing the URL"
-        else
-            echo "Waiting for service to respond... ($i/60, current status: ${http_code})"
-            sleep 5
-        fi
-    done
-
     echo ""
     echo "Querybook installed successfully!"
     echo "Access URL: https://${QUERYBOOK_HOST}"

querybook/querybook-values.gomplate.yaml

@@ -15,13 +15,15 @@ worker:
     pullPolicy: IfNotPresent
     tag: latest
 {{- end }}
+  # Resource limits (based on Goldilocks/VPA recommendations, rounded to clean values)
+  # VPA recommendations: CPU 15m, Memory 2.8Gi
   resources:
     requests:
-      memory: 1Gi
-      cpu: 700m
+      cpu: 25m
+      memory: 3Gi
     limits:
-      memory: 2Gi
-      cpu: 1
+      cpu: 500m
+      memory: 6Gi
 
 # Scheduler configuration
 scheduler:
@@ -37,13 +39,15 @@ scheduler:
     pullPolicy: IfNotPresent
     tag: latest
 {{- end }}
+  # Resource limits (based on Goldilocks/VPA recommendations, rounded to clean values)
+  # VPA recommendations: CPU 15m, Memory 194Mi
   resources:
     requests:
-      memory: 200Mi
-      cpu: 100m
+      cpu: 25m
+      memory: 256Mi
     limits:
-      memory: 300Mi
-      cpu: 200m
+      cpu: 100m
+      memory: 512Mi
 
 # Web server configuration
 web:
@@ -63,13 +67,15 @@ web:
     serviceType: ClusterIP
     servicePort: 80
     containerPort: 10001
+  # Resource limits (based on Goldilocks/VPA recommendations, rounded to clean values)
+  # VPA recommendations: CPU 224m, Memory 215Mi
   resources:
     requests:
-      memory: 1Gi
-      cpu: 500m
+      cpu: 250m
+      memory: 256Mi
     limits:
-      memory: 2Gi
-      cpu: 1
+      cpu: 500m
+      memory: 512Mi
 
   # Custom initContainer to inject Keycloak auth backend
   initContainers:
@@ -84,6 +90,17 @@ web:
           mountPath: /config
         - name: auth-volume
           mountPath: /auth
+      securityContext:
+        allowPrivilegeEscalation: false
+        readOnlyRootFilesystem: false
+        runAsNonRoot: true
+        runAsUser: 1000
+        runAsGroup: 1000
+        seccompProfile:
+          type: RuntimeDefault
+        capabilities:
+          drop:
+            - ALL
 
   # Volume mounts for main container
   volumeMounts:
@@ -115,13 +132,15 @@ redis:
   service:
     serviceType: ClusterIP
     servicePort: 6379
+  # Resource limits (based on Goldilocks/VPA recommendations, rounded to clean values)
+  # VPA recommendations: CPU 15m, Memory 100Mi
   resources:
     requests:
-      memory: 512Mi
-      cpu: 200m
+      cpu: 25m
+      memory: 128Mi
     limits:
-      memory: 1Gi
-      cpu: 500m
+      cpu: 100m
+      memory: 256Mi
 
 # Elasticsearch configuration (use Helm chart's embedded Elasticsearch)
 elasticsearch:
@@ -144,13 +163,15 @@ elasticsearch:
   service:
     serviceType: ClusterIP
     servicePort: 9200
+  # Resource limits (based on Goldilocks/VPA recommendations, rounded to clean values)
+  # VPA recommendations: CPU 78m, Memory 1.7Gi
   resources:
     requests:
+      cpu: 100m
       memory: 2Gi
-      cpu: 500m
     limits:
-      memory: 3Gi
-      cpu: 1
+      cpu: 500m
+      memory: 4Gi
 
 # Ingress configuration
 ingress: