feat(elastic-as-graph): defaulting to elastic in quickstart

datahub-kubernetes/README.md

@@ -34,12 +34,15 @@ The main components are powered by 4 external dependencies:
 - Kafka
 - Local DB (MySQL, Postgres, MariaDB)
 - Search Index (Elasticsearch)
-- Graph Index (Supports only Neo4j)
+- Graph Index (Supports either Neo4j or Elasticsearch)
 
 The dependencies must be deployed before deploying Datahub. We created a separate 
 [chart](https://github.com/linkedin/datahub/tree/master/datahub-kubernetes/prerequisites) 
 for deploying the dependencies with example configuration. They could also be deployed 
-separately on-prem or leveraged as managed services.   
+separately on-prem or leveraged as managed services. To remove your dependency on Neo4j,
+set enabled to false in the `datahub-kubernetes/prerequisites/values.yaml` file.
+Then, override the `graph_service_impl` field in `datahub-kubernetes/datahub/values.yaml` to
+have the value `elasticsearch` instead of `neo4j`.
 
 ## Quickstart
 Assuming kubectl context points to the correct kubernetes cluster, first create kubernetes secrets that contain MySQL and Neo4j passwords. 
@@ -130,5 +133,3 @@ to expose the 9002 port to the public.
 | helm uninstall datahub | Remove DataHub |
 | helm ls | List of Helm charts |
 | helm history | Fetch a release history | 
-
-

docker/README.md

@@ -32,7 +32,7 @@ Dependencies:
 * [Kafka, Zookeeper, and Schema Registry](kafka-setup)
 * [Elasticsearch](elasticsearch-setup)
 * [MySQL](mysql)
-* [Neo4j](neo4j)
+* [(Optional) Neo4j](neo4j)
 
 ### Ingesting demo data.
 

docker/quickstart.sh

@@ -15,4 +15,17 @@ DEFAULT_VERSION=$(echo $TAG_VERSION | sed 's/undefined/head/')
 export DATAHUB_VERSION=${DATAHUB_VERSION:-${DEFAULT_VERSION}}
 
 echo "Quickstarting DataHub: version ${DATAHUB_VERSION}"
-cd $DIR && docker-compose pull && docker-compose -p datahub up
+if docker volume ls | grep -c -q datahub_neo4jdata
+then
+  echo "Datahub Neo4j volume found, starting with neo4j as graph service"
+  cd $DIR && docker-compose pull && docker-compose -p datahub up
+else
+  echo "No Datahub Neo4j volume found, starting with elasticsearch as graph service"
+  cd $DIR && \
+  docker-compose \
+    -f quickstart/docker-compose-without-neo4j.quickstart.yml \
+    pull && \
+  docker-compose -p datahub \
+    -f quickstart/docker-compose-without-neo4j.quickstart.yml \
+    up
+fi

metadata-ingestion/src/datahub/cli/docker.py

@@ -17,11 +17,21 @@
 )
 from datahub.ingestion.run.pipeline import Pipeline
 
-SIMPLE_QUICKSTART_COMPOSE_FILE = "docker/quickstart/docker-compose.quickstart.yml"
+NEO4J_AND_ELASTIC_QUICKSTART_COMPOSE_FILE = (
+    "docker/quickstart/docker-compose.quickstart.yml"
+)
+ELASTIC_QUICKSTART_COMPOSE_FILE = (
+    "docker/quickstart/docker-compose-without-neo4j.quickstart.yml"
+)
 BOOTSTRAP_MCES_FILE = "metadata-ingestion/examples/mce_files/bootstrap_mce.json"
 
 GITHUB_BASE_URL = "https://raw.githubusercontent.com/linkedin/datahub/master"
-GITHUB_QUICKSTART_COMPOSE_URL = f"{GITHUB_BASE_URL}/{SIMPLE_QUICKSTART_COMPOSE_FILE}"
+GITHUB_NEO4J_AND_ELASTIC_QUICKSTART_COMPOSE_URL = (
+    f"{GITHUB_BASE_URL}/{NEO4J_AND_ELASTIC_QUICKSTART_COMPOSE_FILE}"
+)
+GITHUB_ELASTIC_QUICKSTART_COMPOSE_URL = (
+    f"{GITHUB_BASE_URL}/{ELASTIC_QUICKSTART_COMPOSE_FILE}"
+)
 GITHUB_BOOTSTRAP_MCES_URL = f"{GITHUB_BASE_URL}/{BOOTSTRAP_MCES_FILE}"
 
 
@@ -58,6 +68,30 @@ def check() -> None:
     docker_check_impl()
 
 
+def check_neo4j_volume_exists():
+    with get_client_with_error() as (client, error):
+        if error:
+            click.secho(
+                "Docker doesn't seem to be running. Did you start it?", fg="red"
+            )
+            return
+
+        if len(client.volumes.list(filters={"name": "datahub_neo4jdata"})) > 0:
+            click.echo(
+                "Datahub Neo4j volume found, starting with neo4j as graph service.\n"
+                "If you want to run using elastic, run `datahub docker nuke` and re-ingest your data.\n"
+            )
+            return True
+
+        click.echo(
+            "No Datahub Neo4j volume found, starting with elasticsearch as graph service.\n"
+            "To use neo4j as a graph backend, run \n"
+            "`datahub docker quickstart --quickstart-compose-file ./docker/quickstart/docker-compose.quickstart.yml`"
+            "\nfrom the root of the datahub repo\n"
+        )
+        return False
+
+
 @docker.command()
 @click.option(
     "--version",
@@ -115,7 +149,11 @@ def quickstart(
             quickstart_compose_file.append(path)
 
             # Download the quickstart docker-compose file from GitHub.
-            quickstart_download_response = requests.get(GITHUB_QUICKSTART_COMPOSE_URL)
+            quickstart_download_response = requests.get(
+                GITHUB_NEO4J_AND_ELASTIC_QUICKSTART_COMPOSE_URL
+                if check_neo4j_volume_exists()
+                else GITHUB_ELASTIC_QUICKSTART_COMPOSE_URL
+            )
             quickstart_download_response.raise_for_status()
             tmp_file.write(quickstart_download_response.content)
 

metadata-ingestion/src/datahub/cli/docker_check.py

@@ -12,7 +12,6 @@
     "schema-registry",
     "broker",
     "mysql",
-    "neo4j",
     "zookeeper",
     # These two containers are not necessary - only helpful in debugging.
     # "kafka-topics-ui",
@@ -33,6 +32,7 @@
     # We only add this container in some cases, but if it's present, we
     # definitely want to check that it exits properly.
     "mysql-setup",
+    "neo4j",
 ]
 
 # Docker seems to under-report memory allocated, so we also need a bit of buffer to account for it.