feat(kubernetes): add agent execution mode, executor-agent, sidecar refactor, GKE Sandbox, image & deps upgrades

.env.example

@@ -32,6 +32,8 @@ API_KEY_CACHE_TTL=300
 RATE_LIMIT_ENABLED=true
 
 # Redis Configuration
+# Deployment mode: standalone (default), cluster, or sentinel
+REDIS_MODE=standalone
 REDIS_HOST=localhost
 REDIS_PORT=6379
 REDIS_PASSWORD=
@@ -42,6 +44,33 @@ REDIS_MAX_CONNECTIONS=20
 REDIS_SOCKET_TIMEOUT=5
 REDIS_SOCKET_CONNECT_TIMEOUT=5
 
+# Optional key prefix — useful when sharing a Redis instance across environments
+# All keys will be stored as <prefix><key>  (e.g. "prod:sessions:abc")
+REDIS_KEY_PREFIX=
+
+# Redis Cluster Mode (REDIS_MODE=cluster)
+# Comma-separated list of host:port pairs for cluster startup nodes
+# REDIS_CLUSTER_NODES=node1:6379,node2:6379,node3:6379
+
+# Redis Sentinel Mode (REDIS_MODE=sentinel)
+# Comma-separated list of host:port pairs for Sentinel instances
+# REDIS_SENTINEL_NODES=sentinel1:26379,sentinel2:26379,sentinel3:26379
+# REDIS_SENTINEL_MASTER=mymaster
+# REDIS_SENTINEL_PASSWORD=
+
+# Redis TLS/SSL Configuration
+# Required for most managed Redis services (GCP Memorystore, AWS ElastiCache, Azure Cache)
+REDIS_TLS_ENABLED=false
+# REDIS_TLS_CA_CERT_FILE=/path/to/ca.crt
+# REDIS_TLS_CERT_FILE=/path/to/client.crt
+# REDIS_TLS_KEY_FILE=/path/to/client.key
+# REDIS_TLS_INSECURE=false
+# Hostname verification is off by default because managed Redis services
+# and Redis Cluster mode expose node IPs that don't match cert CN/SAN.
+# The CA certificate chain is still fully verified. Enable hostname
+# checking when your Redis server hostnames match certificate CN/SAN.
+# REDIS_TLS_CHECK_HOSTNAME=false
+
 # MinIO/S3 Configuration
 MINIO_ENDPOINT=localhost:9000
 MINIO_ACCESS_KEY=minioadmin
@@ -144,6 +173,37 @@ METRICS_ARCHIVE_RETENTION_DAYS=90
 ENABLE_NETWORK_ISOLATION=true
 ENABLE_FILESYSTEM_ISOLATION=true
 
+# Kubernetes Execution Configuration
+# Execution mode: 'agent' (default, recommended) or 'nsenter' (legacy)
+#   agent:   Executor-agent binary runs inside the main container.
+#            No nsenter, no capabilities, no privilege escalation.
+#            Compatible with GKE Sandbox (gVisor) and restricted Pod Security Standards.
+#   nsenter: Sidecar uses nsenter to enter the main container's mount namespace.
+#            Requires shareProcessNamespace, SYS_PTRACE/SYS_ADMIN/SYS_CHROOT caps,
+#            and allowPrivilegeEscalation: true. NOT compatible with GKE Sandbox.
+K8S_EXECUTION_MODE=agent
+# K8S_EXECUTOR_PORT=9090  # Port for the executor-agent HTTP server (agent mode only)
+
+# Sidecar image — must match the execution mode:
+#   agent mode:   aronmuon/kubecoderun-sidecar-agent:latest   (default)
+#   nsenter mode: aronmuon/kubecoderun-sidecar-nsenter:latest
+# K8S_SIDECAR_IMAGE=aronmuon/kubecoderun-sidecar-agent:latest
+
+# Image pull policy for execution pods (Always, IfNotPresent, Never)
+# K8S_IMAGE_PULL_POLICY=Always
+
+# Image pull secrets for private container registries (comma-separated secret names)
+# These Kubernetes secrets must already exist in the execution namespace.
+# Leave empty or unset if not using private registries.
+# K8S_IMAGE_PULL_SECRETS=my-registry-secret,another-secret
+
+# GKE Sandbox (gVisor) Configuration
+# Requires K8S_EXECUTION_MODE=agent (nsenter is incompatible with gVisor)
+# GKE_SANDBOX_ENABLED=false
+# GKE_SANDBOX_RUNTIME_CLASS=gvisor
+# GKE_SANDBOX_NODE_SELECTOR={}
+# GKE_SANDBOX_CUSTOM_TOLERATIONS=[]
+
 # WAN Network Access Configuration
 # When enabled, execution containers can access the public internet
 # but are blocked from accessing host, other containers, and private networks

.github/workflows/docker-publish.yml

@@ -145,7 +145,7 @@ jobs:
     uses: ./.github/workflows/docker-build-reusable.yml
     secrets: inherit
     with:
-      image_name: kubecoderun-sidecar
+      image_name: kubecoderun-sidecar-agent
       dockerfile: docker/sidecar/Dockerfile
       context: docker/sidecar
       image_tag: ${{ needs.changes.outputs.image_tag }}
@@ -344,7 +344,7 @@ jobs:
     uses: ./.github/workflows/docker-retag-reusable.yml
     secrets: inherit
     with:
-      image_name: kubecoderun-sidecar
+      image_name: kubecoderun-sidecar-agent
       new_tag: ${{ needs.changes.outputs.image_tag }}
       previous_tag: ${{ needs.changes.outputs.previous_tag }}
 

.gitignore

@@ -200,3 +200,5 @@ config/local.py
 
 # Hatch auto-generated version file
 _version.py
+
+.pdm-python

docker-compose.redis-cluster-tls.yml

@@ -0,0 +1,231 @@
+# Redis Cluster with TLS for integration testing
+#
+# This mimics a production GCP Memorystore Redis Cluster setup:
+# - 6-node cluster (3 masters + 3 replicas) with TLS enabled
+# - No authentication (no password)
+# - Server-side TLS with CA verification (no mutual TLS / no client certs)
+# - Accessible on localhost ports 6380-6385 (TLS)
+#
+# Usage:
+#   docker compose -f docker-compose.redis-cluster-tls.yml up -d
+#
+# Test with:
+#   redis-cli -c -p 6380 --tls --cacert tests/tls-certs/ca.crt CLUSTER INFO
+
+services:
+  redis-tls-node-0:
+    image: redis:7-alpine
+    container_name: redis-tls-cluster-0
+    ports:
+      - "127.0.0.1:6380:6380"
+      - "127.0.0.1:16380:16380"
+    volumes:
+      - redis-tls-cluster-0:/data
+      - ./tests/tls-certs:/tls:ro
+    command: >
+      redis-server
+      --port 0
+      --tls-port 6380
+      --tls-cert-file /tls/redis.crt
+      --tls-key-file /tls/redis.key
+      --tls-ca-cert-file /tls/ca.crt
+      --tls-auth-clients no
+      --tls-replication yes
+      --cluster-enabled yes
+      --cluster-config-file nodes.conf
+      --cluster-node-timeout 5000
+      --appendonly yes
+      --bind 0.0.0.0
+      --protected-mode no
+    healthcheck:
+      test: ["CMD", "redis-cli", "-p", "6380", "--tls", "--cert", "/tls/redis.crt", "--key", "/tls/redis.key", "--cacert", "/tls/ca.crt", "ping"]
+      interval: 5s
+      timeout: 3s
+      retries: 10
+
+  redis-tls-node-1:
+    image: redis:7-alpine
+    container_name: redis-tls-cluster-1
+    ports:
+      - "127.0.0.1:6381:6381"
+      - "127.0.0.1:16381:16381"
+    volumes:
+      - redis-tls-cluster-1:/data
+      - ./tests/tls-certs:/tls:ro
+    command: >
+      redis-server
+      --port 0
+      --tls-port 6381
+      --tls-cert-file /tls/redis.crt
+      --tls-key-file /tls/redis.key
+      --tls-ca-cert-file /tls/ca.crt
+      --tls-auth-clients no
+      --tls-replication yes
+      --cluster-enabled yes
+      --cluster-config-file nodes.conf
+      --cluster-node-timeout 5000
+      --appendonly yes
+      --bind 0.0.0.0
+      --protected-mode no
+    healthcheck:
+      test: ["CMD", "redis-cli", "-p", "6381", "--tls", "--cert", "/tls/redis.crt", "--key", "/tls/redis.key", "--cacert", "/tls/ca.crt", "ping"]
+      interval: 5s
+      timeout: 3s
+      retries: 10
+
+  redis-tls-node-2:
+    image: redis:7-alpine
+    container_name: redis-tls-cluster-2
+    ports:
+      - "127.0.0.1:6382:6382"
+      - "127.0.0.1:16382:16382"
+    volumes:
+      - redis-tls-cluster-2:/data
+      - ./tests/tls-certs:/tls:ro
+    command: >
+      redis-server
+      --port 0
+      --tls-port 6382
+      --tls-cert-file /tls/redis.crt
+      --tls-key-file /tls/redis.key
+      --tls-ca-cert-file /tls/ca.crt
+      --tls-auth-clients no
+      --tls-replication yes
+      --cluster-enabled yes
+      --cluster-config-file nodes.conf
+      --cluster-node-timeout 5000
+      --appendonly yes
+      --bind 0.0.0.0
+      --protected-mode no
+    healthcheck:
+      test: ["CMD", "redis-cli", "-p", "6382", "--tls", "--cert", "/tls/redis.crt", "--key", "/tls/redis.key", "--cacert", "/tls/ca.crt", "ping"]
+      interval: 5s
+      timeout: 3s
+      retries: 10
+
+  redis-tls-node-3:
+    image: redis:7-alpine
+    container_name: redis-tls-cluster-3
+    ports:
+      - "127.0.0.1:6383:6383"
+      - "127.0.0.1:16383:16383"
+    volumes:
+      - redis-tls-cluster-3:/data
+      - ./tests/tls-certs:/tls:ro
+    command: >
+      redis-server
+      --port 0
+      --tls-port 6383
+      --tls-cert-file /tls/redis.crt
+      --tls-key-file /tls/redis.key
+      --tls-ca-cert-file /tls/ca.crt
+      --tls-auth-clients no
+      --tls-replication yes
+      --cluster-enabled yes
+      --cluster-config-file nodes.conf
+      --cluster-node-timeout 5000
+      --appendonly yes
+      --bind 0.0.0.0
+      --protected-mode no
+    healthcheck:
+      test: ["CMD", "redis-cli", "-p", "6383", "--tls", "--cert", "/tls/redis.crt", "--key", "/tls/redis.key", "--cacert", "/tls/ca.crt", "ping"]
+      interval: 5s
+      timeout: 3s
+      retries: 10
+
+  redis-tls-node-4:
+    image: redis:7-alpine
+    container_name: redis-tls-cluster-4
+    ports:
+      - "127.0.0.1:6384:6384"
+      - "127.0.0.1:16384:16384"
+    volumes:
+      - redis-tls-cluster-4:/data
+      - ./tests/tls-certs:/tls:ro
+    command: >
+      redis-server
+      --port 0
+      --tls-port 6384
+      --tls-cert-file /tls/redis.crt
+      --tls-key-file /tls/redis.key
+      --tls-ca-cert-file /tls/ca.crt
+      --tls-auth-clients no
+      --tls-replication yes
+      --cluster-enabled yes
+      --cluster-config-file nodes.conf
+      --cluster-node-timeout 5000
+      --appendonly yes
+      --bind 0.0.0.0
+      --protected-mode no
+    healthcheck:
+      test: ["CMD", "redis-cli", "-p", "6384", "--tls", "--cert", "/tls/redis.crt", "--key", "/tls/redis.key", "--cacert", "/tls/ca.crt", "ping"]
+      interval: 5s
+      timeout: 3s
+      retries: 10
+
+  redis-tls-node-5:
+    image: redis:7-alpine
+    container_name: redis-tls-cluster-5
+    ports:
+      - "127.0.0.1:6385:6385"
+      - "127.0.0.1:16385:16385"
+    volumes:
+      - redis-tls-cluster-5:/data
+      - ./tests/tls-certs:/tls:ro
+    command: >
+      redis-server
+      --port 0
+      --tls-port 6385
+      --tls-cert-file /tls/redis.crt
+      --tls-key-file /tls/redis.key
+      --tls-ca-cert-file /tls/ca.crt
+      --tls-auth-clients no
+      --tls-replication yes
+      --cluster-enabled yes
+      --cluster-config-file nodes.conf
+      --cluster-node-timeout 5000
+      --appendonly yes
+      --bind 0.0.0.0
+      --protected-mode no
+    healthcheck:
+      test: ["CMD", "redis-cli", "-p", "6385", "--tls", "--cert", "/tls/redis.crt", "--key", "/tls/redis.key", "--cacert", "/tls/ca.crt", "ping"]
+      interval: 5s
+      timeout: 3s
+      retries: 10
+
+  # Initializer: creates TLS cluster from the 6 nodes
+  redis-tls-cluster-init:
+    image: redis:7-alpine
+    container_name: redis-tls-cluster-init
+    volumes:
+      - ./tests/tls-certs:/tls:ro
+    depends_on:
+      redis-tls-node-0:
+        condition: service_healthy
+      redis-tls-node-1:
+        condition: service_healthy
+      redis-tls-node-2:
+        condition: service_healthy
+      redis-tls-node-3:
+        condition: service_healthy
+      redis-tls-node-4:
+        condition: service_healthy
+      redis-tls-node-5:
+        condition: service_healthy
+    restart: "no"
+    entrypoint:
+      - sh
+      - -c
+      - |
+        echo 'Creating Redis TLS Cluster...' &&
+        redis-cli --cluster create redis-tls-node-0:6380 redis-tls-node-1:6381 redis-tls-node-2:6382 redis-tls-node-3:6383 redis-tls-node-4:6384 redis-tls-node-5:6385 --cluster-replicas 1 --cluster-yes --tls --cert /tls/redis.crt --key /tls/redis.key --cacert /tls/ca.crt &&
+        echo 'Redis TLS Cluster created successfully' &&
+        redis-cli -h redis-tls-node-0 -p 6380 --tls --cert /tls/redis.crt --key /tls/redis.key --cacert /tls/ca.crt CLUSTER INFO
+
+volumes:
+  redis-tls-cluster-0:
+  redis-tls-cluster-1:
+  redis-tls-cluster-2:
+  redis-tls-cluster-3:
+  redis-tls-cluster-4:
+  redis-tls-cluster-5: