In recent years, MongoDB has become a popular online transactional processing
(OLTP) option because of its scalability and the flexibility inherent in NoSQL databases. However, as data scales, some developers feel the need to migrate to an online analytical processing (OLAP) specific platform like Apache Druid, for more advanced analytical capabilities beyond what MongoDB can offer. This blog will guide readers on migrating data from MongoDB to Druid, including the steps and code to connect to MongoDB, extract the data, and then ingest it into Druid.
Druid is specifically designed for high-performance analytical queries. It stores data in a columnar format but what sets Druid apart is its indexing engine and segmented data storage which tremendously boosts query performance. The indexing allows Druid to swiftly filter and aggregate data at the segment level. Consequently, Druid can fetch only the pertinent data for a specific query instead of scanning the entire dataset. This method considerably cuts down the I/O operations required to pull data from storage, leading to faster query results.
The Data Migration Process
To execute the steps outlined in this blog, you will need the following prerequisites:
- Apache Druid installed
- Account on MongoDB Atlas
Extracting Data from MongoDBThe Python script below is designed to fetch data from a MongoDB database and store the retrieved records in a CSV file format. The script loads MongoDB connection configurations from the database_configs.json file using the load_config function. Then the connect_to_mongodb function establishes a connection to the MongoDB instance, utilizing the credentials and connection details specified in the configuration. With the connection established, the fetch_documents function is invoked to pull all documents from the customer_data collection in the my_sample database. These documents are stored in a CSV file with the save_results_to_csv function. If any exceptions or errors arise during the process, relevant error messages are shown to the user. After the data retrieval and storage processes are completed, the script calls the create_ingest_spec function from the mongodb_ingest module which will be shown later.
import csv
import json
import pymongo
import sys
from pymongo import MongoClient
from pymongo.server_api import ServerApi
from mongodb_ingest import create_ingest_spec
def load_config(file_path='database_configs.json'):
"""Load MongoDB configuration from a JSON file."""
with open(file_path, 'r') as file:
return json.load(file)
def connect_to_mongodb(config):
"""Connect to MongoDB using the provided configuration."""
user = config['mongodb']['user']
password = config['mongodb']['password']
cluster = config['mongodb']['cluster']
uri = f"mongodb+srv://{user}:{password}{cluster}"
client = MongoClient(uri, server_api=ServerApi('1'))
return client
def fetch_documents(client, db_name, collection_name):
"""Fetch documents from the specified database and collection."""
db = client[db_name]
collection = db[collection_name]
return list(collection.find())
def save_results_to_csv(documents, output_file="data/mongodb_data.csv"):
"""Save the result set to a CSV file."""
with open(output_file, 'w', newline='') as file:
writer = csv.writer(file)
if documents:
writer.writerow(documents[0].keys())
for document in documents:
writer.writerow(document.values())
def get_data():
try:
config = load_config()
with connect_to_mongodb(config) as client:
print("Connected to MongoDB!")
results = fetch_documents(client, 'my_sample', 'customer_data')
save_results_to_csv(results)
except pymongo.errors.ConfigurationError:
print("Invalid URI host error. Check the Atlas host name in your connection string.")
except Exception as e:
print(f"An unexpected error occurred: {e}")
sys.exit(1)
if __name__ == "__main__":
get_data()
create_ingest_spec()Note that the load_config function uses a config file to store MongoDB connection data.
Below is an example of the `database_configs.json` that stores the database configuration details that are used in the script above.
{
"MongoDB": {
"user": "rjsample",
"password": "xxxxxxxxx",
"account": "xxxxxxx.us-east-1",
"database": "DEMO_SAMPLE_DATA"
}
}Loading Data into Druid
For this batch upload process, the script below constructs an ingestion specification for Druid using the construct_ingestion_spec function. This function takes the name of the data source and the directory where the CSV data is located. The post_ingest_spec_to_druid function is utilized to send the specification to Druid. This function employs the requests library to dispatch the ingestion specification via an HTTP POST request. If any errors arise during this process, they are captured and presented to the user. For added convenience, the script comes with default values for the data source, directory, and Druid host which can be overridden by passing alternate parameters to the create_ingest_spec function.
import json
import requests
def construct_ingestion_spec(DATA_SOURCE, BASE_DIR):
"""Construct the ingestion spec based on the data source and directory."""
return {
"type": "index_parallel",
"spec": {
"ioConfig": {
"type": "index_parallel",
"inputSource": {
"type": "local",
"baseDir": BASE_DIR,
"filter": f"{DATA_SOURCE}.csv"
},
"inputFormat": {
"type": "csv",
"findColumnsFromHeader": True
}
},
"tuningConfig": {
"type": "index_parallel",
"partitionsSpec": {
"type": "dynamic"
}
},
"dataSchema": {
"dataSource": DATA_SOURCE,
"timestampSpec": {
"column": "CREATED_AT",
"format": "auto"
},
"dimensionsSpec": {
"useSchemaDiscovery": True,
"dimensionExclusions": []
},
"granularitySpec": {
"queryGranularity": "none",
"rollup": False,
"segmentGranularity": "hour"
}
}
}
}
def post_ingest_spec_to_druid(spec, DRUID_HOST):
"""Post the ingestion spec to Druid and return the response."""
headers = {'Content-Type': 'application/json'}
response = requests.post(DRUID_HOST, json.dumps(spec), headers=headers)
return response
def create_ingest_spec(DATA_SOURCE='mongodb_data',
BASE_DIR='/Users/rick/IdeaProjects/CodeProjects/druid_data_integrations/data',
DRUID_HOST='http://localhost:8081/druid/indexer/v1/task'):
"""Create the ingestion spec and post it to Druid."""
spec = construct_ingestion_spec(DATA_SOURCE, BASE_DIR)
print(f'Creating .csv ingestion spec for {DATA_SOURCE}')
try:
response = post_ingest_spec_to_druid(spec, DRUID_HOST)
response.raise_for_status()
print(response.text)
except requests.HTTPError:
print(f"HTTP error occurred: {response.text}")
except Exception as e:
print(f"An error occurred: {e}")
if __name__ == "__main__":
create_ingest_spec()After the primary data ingestion, subsequent updates from MongoDB can be efficiently managed using a change data capture (CDC) approach. This method exploits the LAST_UPDATED timestamp to pinpoint new or modified records. Then, either data loading techniques are employed to append or refresh the records, or various ELT tools are utilized. For example:
- MongoDB Change Streams: MongoDB boasts a feature termed “Change Streams” which lets users track and respond to data modifications in real-time. Leveraging MongoDB Change Streams, you can pinpoint altered data and then employ ETL tools or bespoke scripts to channel these modifications into Druid.
- ETL Tools: Established tools such as Apache NiFi, Talend, or StreamSets offer robust capabilities to design workflows that can capture, process, and transfer alterations from MongoDB straight to Druid.
- Custom Scripts: If your changes exhibit a unique complexity or sheer volume, it might be worthwhile to develop custom scripts. These scripts can be scheduled to routinely check MongoDB for modifications and subsequently relay these updates to Druid.
- Third-party Solutions: The market offers specialized third-party solutions and services with inherent CDC features. These can be tailored to bridge MongoDB and Druid, offering a hands-off, managed approach for those who prefer it.
Sample Query
To verify that the data was ingested, select the query interface from within the Druid console and execute SQL similar to the example below:
Summary
MongoDB is one of the more popular NoSQL databases among developers. However, for fast analytics on large datasets, moving data to Druid may offer more efficient data analysis and quicker dynamic visualizations. Druid’s design philosophy is centered around delivering swift analytical queries. It employs a columnar data structure, liberal indexing, and segment-oriented storage, which form the foundation of its superior performance. These design choices ensure that only the relevant data chunks are accessed during queries, optimizing resource usage. Additionally, Druid’s native support for real-time data streams establishes it as a top choice for real-time analytics. By adhering to the steps outlined in this blog, you can harness your data more effectively for near real-time analytics and ensure swifter data-based decision-making.
About the Author
Rick Jacobs is a Senior Technical Product Marketing Manager at Imply. His varied background includes experience at IBM, Cloudera, and Couchbase. He has over 20 years of technology experience garnered from serving in development, consulting, data science, sales engineering, and other roles. He holds several academic degrees including an MS in Computational Science from George Mason University. When not working on technology, Rick is trying to learn Spanish and pursuing his dream of becoming a beach bum.
Architecture
Deployment
Ingestion
Modeling
Operations
Development
