Monthly Archives: September 2021

Azure Analytics Using Azure Delta Lake(Databricks): A Step-by-Step Guide

If data sources are installed at a long distance from the stakeholders who control the operation, timely notifications are required.
Agriculture, mining, power generation, oil and gas, and other industries can all benefit

Technologies used: Azure Databricks, Python, pyspark, Scala, Power BI

Data sources: Sensors – like heat, humidity, water leakage, electric meters, drones sending data to iot hub.

Solution description: Data from IOT sources are loaded into ADLS Gen2. A checkpoint is created, and autoloader is configured to load and maintain the data. Lastly Azure data bricks bronze, silver & gold tables are created which supplies data to power BI dashboards based on business rules.


Solution Details:

Step 1: Initial configuration, autoloader & load data

Data from various sources are periodically copied to iot-hub, Now, once file is loaded, autoloader tool ingests this data to azure data lake storage gen2 (ADLS Gen2) databricks mount.

Step 2: Create Delta Lake Bronze table

All raw data in csv format can be ingested to bronze table. Basically, it is complete load of all data received.

Step 3: Data processing in Silver Table

This is one of the major steps and in this solution, data processing is performed and cleaned data is saved in silver table.

Step 4: Data processing in Gold Table

In this step, as per business requirements, business logic is applied, and alerts columns are created. This is applied on current load.

Step 5: Dashboard (Power BI):

Gold table is source for Power BI dashboard. This dashboard can help user to sort the existing alert get more details about it. User can also see different analytics on the dashboard.

Based on business mandate, email notification is sent to stake holders

Code Details:

Initial Connection

import json

import csv

import pyspark

from pyspark.sql import SparkSession

configs = {“”: “OAuth”,

          “”: “org.apache.hadoop.fs.azurebfs.oauth2.ClientCredsTokenProvider”,

          “”: dbutils.secrets.get(scope = “ Scope “, key = “<your ID >”),

          “”: dbutils.secrets.get(scope = “Scope“, key = “Your Key“),      

          “”: “—2222-497b-xxxx-XXXXXX/oauth2/token&#8221; }

adlsPath = “<<my-name>@<>>” # Source

mountPoint = “<Mount point Path>” # Upload Path – # Destination

print(f”adls Path: {adlsPath} “)

print(f”Client Secret : {configs} “)

display(“Mount point Path “))

Autoloader configuration

# Run the following code for autoloader configuration

checkpoint_path = ‘/Mount point Path/_checkpointname/’

import json
import csv
from pyspark.sql.functions import input_file_name
cloudfile = {
  "cloudFiles.format": "csv",
  "cloudFiles.schemaEvolutionMode": "addNewColumns",
  "cloudFiles.inferColumnTypes": "true",
  "cloudFiles.includeExistingFiles": "true",
  "cloudFiles.allowOverwrites": "false",
  "cloudFiles.schemaLocation": "/Mount point Path /_checkpointname/", 
# Set up the stream to begin reading incoming files from the Mount point Path location.
df = (spark.readStream.format("cloudFiles").options(**cloudfile).load('/Mount point Path/StreamingLog) withColumn("filePath",input_file_name()))  # this adds a column with file name 

Create Dataframe using new uploaded data

Here checkpoint is tracking or keeping a record of all the files that are uploaded to mountpoint. newly uploaded files those don’t have any record in checkpoint is loaded at DELTALAKE_BRONZE_PATH

# Start the stream & write the data

DELTALAKE_BRONZE_PATH = "dbfs:/FileStore/Bronze_StreamingLog "
.option("checkpointLocation", "/Mount point Path /_checkpointname/")\

Create Delta Lake Bronze table

# Register the SQL table in the database
# Read the table
streaminglog_stats ="delta").load(DELTALAKE_BRONZE_PATH)

Silver & Gold Table: Once data is uploaded into the bronze table, all data cleaning and ETL can be performed on it and clean data can be saved into the silver table.

# Configure destination path
DELTALAKE_SILVER_PATH = "dbfs:/FileStore/Silver_StreamingLog "
# Write out the table
# Register the SQL table in the database
spark.sql("CREATE TABLE if not exists <Silver_Tablename> USING DELTA LOCATION '" + DELTALAKE_SILVER_PATH + "'")
# Read the table
streaminglog _stats ="delta").load(DELTALAKE_SILVER_PATH)
display(streaminglog _stats)

After formation of silver table, all business rules are applied, and gold table is created. This is the source for all reporting, dash boards and reporting tools, (Power BI) in our case.