feat(opensearch): Add OpenSearch in docker compose (#8611)

Author: acaprau

.github/workflows/pr-external-dependency-unit-tests.yml

@@ -45,9 +45,6 @@ env:
   # TODO: debug why this is failing and enable
   CODE_INTERPRETER_BASE_URL: http://localhost:8000
 
-  # OpenSearch
-  OPENSEARCH_ADMIN_PASSWORD: "StrongPassword123!"
-
 jobs:
   discover-test-dirs:
     # NOTE: Github-hosted runners have about 20s faster queue times and are preferred here.
@@ -118,9 +115,10 @@ jobs:
       - name: Create .env file for Docker Compose
         run: |
           cat <<EOF > deployment/docker_compose/.env
-          COMPOSE_PROFILES=s3-filestore
+          COMPOSE_PROFILES=s3-filestore,opensearch-enabled
           CODE_INTERPRETER_BETA_ENABLED=true
           DISABLE_TELEMETRY=true
+          OPENSEARCH_FOR_ONYX_ENABLED=true
           EOF
 
       - name: Set up Standard Dependencies
@@ -129,7 +127,6 @@ jobs:
           docker compose \
             -f docker-compose.yml \
             -f docker-compose.dev.yml \
-            -f docker-compose.opensearch.yml \
             up -d \
             minio \
             relational_db \

backend/onyx/configs/app_configs.py

@@ -251,7 +251,9 @@
     os.environ.get("DEFAULT_OPENSEARCH_QUERY_TIMEOUT_S") or 50
 )
 OPENSEARCH_ADMIN_USERNAME = os.environ.get("OPENSEARCH_ADMIN_USERNAME", "admin")
-OPENSEARCH_ADMIN_PASSWORD = os.environ.get("OPENSEARCH_ADMIN_PASSWORD", "")
+OPENSEARCH_ADMIN_PASSWORD = os.environ.get(
+    "OPENSEARCH_ADMIN_PASSWORD", "StrongPassword123!"
+)
 USING_AWS_MANAGED_OPENSEARCH = (
     os.environ.get("USING_AWS_MANAGED_OPENSEARCH", "").lower() == "true"
 )
@@ -282,6 +284,9 @@
 ENABLE_OPENSEARCH_INDEXING_FOR_ONYX = (
     os.environ.get("ENABLE_OPENSEARCH_INDEXING_FOR_ONYX", "").lower() == "true"
 )
+# NOTE: This effectively does nothing anymore, admins can now toggle whether
+# retrieval is through OpenSearch. This value is only used as a final fallback
+# in case that doesn't work for whatever reason.
 # Given that the "base" config above is true, this enables whether we want to
 # retrieve from OpenSearch or Vespa. We want to be able to quickly toggle this
 # in the event we see issues with OpenSearch retrieval in our dev environments.

deployment/docker_compose/docker-compose.dev.yml

@@ -35,6 +35,10 @@ services:
       retries: 3
       start_period: 30s
 
+  opensearch:
+    ports:
+      - "9200:9200"
+
   inference_model_server:
     ports:
       - "9000:9000"

deployment/docker_compose/docker-compose.opensearch.yml

@@ -76,6 +76,9 @@ services:
       - FILE_STORE_BACKEND=${FILE_STORE_BACKEND:-s3}
       - POSTGRES_HOST=${POSTGRES_HOST:-relational_db}
       - VESPA_HOST=${VESPA_HOST:-index}
+      - OPENSEARCH_HOST=${OPENSEARCH_HOST:-opensearch}
+      - OPENSEARCH_ADMIN_PASSWORD=${OPENSEARCH_ADMIN_PASSWORD:-StrongPassword123!}
+      - ENABLE_OPENSEARCH_INDEXING_FOR_ONYX=${OPENSEARCH_FOR_ONYX_ENABLED:-false}
       - REDIS_HOST=${REDIS_HOST:-cache}
       - MODEL_SERVER_HOST=${MODEL_SERVER_HOST:-inference_model_server}
       - S3_ENDPOINT_URL=${S3_ENDPOINT_URL:-http://minio:9000}
@@ -147,6 +150,9 @@ services:
       - FILE_STORE_BACKEND=${FILE_STORE_BACKEND:-s3}
       - POSTGRES_HOST=${POSTGRES_HOST:-relational_db}
       - VESPA_HOST=${VESPA_HOST:-index}
+      - OPENSEARCH_HOST=${OPENSEARCH_HOST:-opensearch}
+      - OPENSEARCH_ADMIN_PASSWORD=${OPENSEARCH_ADMIN_PASSWORD:-StrongPassword123!}
+      - ENABLE_OPENSEARCH_INDEXING_FOR_ONYX=${OPENSEARCH_FOR_ONYX_ENABLED:-false}
       - REDIS_HOST=${REDIS_HOST:-cache}
       - MODEL_SERVER_HOST=${MODEL_SERVER_HOST:-inference_model_server}
       - INDEXING_MODEL_SERVER_HOST=${INDEXING_MODEL_SERVER_HOST:-indexing_model_server}
@@ -395,6 +401,45 @@ services:
         max-size: "50m"
         max-file: "6"
 
+  opensearch:
+    image: opensearchproject/opensearch:3.4.0
+    restart: unless-stopped
+    # Controls whether this service runs. In order to enable it, add
+    # opensearch-enabled to COMPOSE_PROFILES in the environment for this
+    # docker-compose.
+    profiles: ["opensearch-enabled"]
+    environment:
+      # We need discovery.type=single-node so that OpenSearch doesn't try
+      # forming a cluster and waiting for other nodes to become live.
+      - discovery.type=single-node
+      - OPENSEARCH_INITIAL_ADMIN_PASSWORD=${OPENSEARCH_ADMIN_PASSWORD:-StrongPassword123!}
+      # This and the JVM config below come from the example in https://docs.opensearch.org/latest/install-and-configure/install-opensearch/docker/
+      # We do this to avoid unstable performance from page swaps.
+      - bootstrap.memory_lock=true # Disable JVM heap memory swapping.
+      # Java heap should be ~50% of memory limit. For now we assume a limit of
+      # 2g although in practice the container can request more than this.
+      # See https://opster.com/guides/opensearch/opensearch-basics/opensearch-heap-size-usage-and-jvm-garbage-collection/
+      # Xms is the starting size, Xmx is the maximum size. These should be the
+      # same.
+      - "OPENSEARCH_JAVA_OPTS=-Xms1g -Xmx1g"
+    volumes:
+      - opensearch-data:/usr/share/opensearch/data
+    # These come from the example in https://docs.opensearch.org/latest/install-and-configure/install-opensearch/docker/
+    ulimits:
+      # Similarly to bootstrap.memory_lock, we don't want to impose limits on
+      # how much memory a process can lock from being swapped.
+      memlock:
+        soft: -1 # Set memlock to unlimited (no soft or hard limit).
+        hard: -1
+      nofile:
+        soft: 65536 # Maximum number of open files for the opensearch user - set to at least 65536.
+        hard: 65536
+    logging:
+      driver: json-file
+      options:
+        max-size: "50m"
+        max-file: "6"
+
   nginx:
     image: nginx:1.25.5-alpine
     restart: unless-stopped
@@ -535,3 +580,5 @@ volumes:
   indexing_model_server_logs:
   # Shared volume for persistent document storage (Craft file-system mode)
   file-system:
+  # Persistent data for OpenSearch.
+  opensearch-data:

deployment/docker_compose/docker-compose.yml

@@ -64,9 +64,13 @@ POSTGRES_PASSWORD=password
 
 ## File Store Backend: "s3" (default, uses MinIO) or "postgres" (no extra services needed)
 ## COMPOSE_PROFILES activates the MinIO service. To use PostgreSQL file storage instead,
-## comment out COMPOSE_PROFILES and set FILE_STORE_BACKEND=postgres.
+## remove s3-filestore from COMPOSE_PROFILES and set FILE_STORE_BACKEND=postgres.
 COMPOSE_PROFILES=s3-filestore
 FILE_STORE_BACKEND=s3
+## Settings for enabling OpenSearch. Uncomment these and comment out
+## COMPOSE_PROFILES above.
+# COMPOSE_PROFILES=s3-filestore,opensearch-enabled
+# OPENSEARCH_FOR_ONYX_ENABLED=true
 
 ## MinIO/S3 Configuration (only needed when FILE_STORE_BACKEND=s3)
 S3_ENDPOINT_URL=http://minio:9000