HIVE-29709: Add health check to docker

packaging/src/docker/docker-compose.yml

@@ -32,11 +32,17 @@ services:
       - hive-db:/var/lib/postgresql
     networks:
       - hive
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U hive -d metastore_db"]
+      interval: 5s
+      timeout: 3s
+      retries: 30
 
   metastore:
     image: apache/hive:${HIVE_VERSION}
     depends_on:
-      - postgres
+      postgres:
+        condition: service_healthy
     restart: unless-stopped
     container_name: metastore
     hostname: metastore
@@ -67,11 +73,18 @@ services:
         - ./jars:/tmp/ext-jars:ro
     networks:
       - hive
+    healthcheck:
+      test: ["CMD-SHELL", "bash -c 'echo > /dev/tcp/localhost/9083'"]
+      interval: 5s
+      timeout: 5s
+      retries: 30
+      start_period: 30s
 
   hiveserver2:
     image: apache/hive:${HIVE_VERSION}
     depends_on:
-      - metastore
+      metastore:
+        condition: service_healthy
     restart: unless-stopped
     container_name: hiveserver2
     hostname: hiveserver2
@@ -125,6 +138,12 @@ services:
       - ./jars:/tmp/ext-jars:ro
     networks:
       - hive
+    healthcheck:
+      test: ["CMD-SHELL", "bash -c 'echo > /dev/tcp/localhost/10000'"]
+      interval: 10s
+      timeout: 10s
+      retries: 30
+      start_period: 40s
 
   zookeeper:
     profiles:
@@ -141,6 +160,11 @@ services:
       - zookeeper_data:/data
       - zookeeper_datalog:/datalog
       - zookeeper_logs:/logs
+    healthcheck:
+      test: ["CMD-SHELL", "bash -c 'echo ruok > /dev/tcp/localhost/2181'"]
+      interval: 5s
+      timeout: 3s
+      retries: 30
 
   tezam:
     profiles:
@@ -149,7 +173,8 @@ services:
     container_name: tezam
     hostname: tezam
     depends_on:
-      - zookeeper
+      zookeeper:
+        condition: service_healthy
     restart: on-failure:3
     environment:
       USER: hive
@@ -177,7 +202,8 @@ services:
       - llap
     image: apache/hive:${HIVE_VERSION}
     depends_on:
-      - zookeeper
+      zookeeper:
+        condition: service_healthy
     restart: unless-stopped
     environment:
       USER: hive
@@ -205,6 +231,12 @@ services:
       - scratch:/opt/hive/scratch
     networks:
       - hive
+    healthcheck:
+      test: ["CMD-SHELL", "bash -c 'echo > /dev/tcp/localhost/15001'"]
+      interval: 5s
+      timeout: 5s
+      retries: 30
+      start_period: 20s
 
 volumes:
   hive-db:

packaging/src/docker/start-hive.sh

@@ -58,4 +58,6 @@ export COMPOSE_FILE="$COMPOSE_FILES"
 
 echo "Starting Hive cluster (mode=$HIVE_EXECUTION_MODE)"
 
-docker compose $PROFILE up -d $SCALE
+docker compose $PROFILE up -d --wait $SCALE
+
+echo "Hive cluster is ready."