HintsToday

Here’s a crisp explanation of core technical terms in Azure Databricks, tailored for interviews and hands-on clarity:

🚀 Databricks Key Technical Terms

🧭 Workspace

UI and environment where users organize notebooks, jobs, data, repos, and libraries.

• Like a folder system for all Databricks assets.
• Shared across users for collaboration.
• Can link to GitHub/DevOps via Repos tab.

⚙️ Cluster

A Spark compute environment managed by Databricks.

• Has one Driver and multiple Workers
• Types:
  • Interactive: For dev & notebooks
  • Job: For scheduled/automated pipelines
• You can configure autoscaling, libraries, runtime version, spot vs on-demand instances.

🧱 DBFS (Databricks File System)

Databricks-managed distributed storage layer on top of your cloud (e.g., ADLS/Blob in Azure).

• Path: /dbfs/...
• Upload files via UI or code.
• Use for temporary or staging data.
• Example: spark.read.csv(\"/dbfs/tmp/data.csv\")

📓 Notebooks

Interactive code documents in Databricks.

• Support Python, SQL, Scala, R, Markdown
• Execute code per-cell
• Track versions, output, and visualizations
• Can be parameterized for jobs and workflows

🛠️ Databricks Runtime

Pre-configured Spark environment with optimized libraries and performance tuning.

• Comes in various types:
  • Standard
  • ML (with scikit-learn, TensorFlow, etc.)
  • Photon (for SQL performance)
  • Delta (for Lakehouse support)
• Versioned (e.g., 13.3 LTS, 14.0 ML)

📦 Libraries

External packages you can install on a cluster.

• Types:
  • Maven: For Scala/Java
  • PyPI: Python packages like pandas, nltk
  • Jar/Egg/Wheel: Custom uploads
• Can be cluster-scoped or notebook-scoped
• Installed via UI or %pip install in notebook

🚗 Driver vs Worker

📌 The Driver runs on 1 machine, and Workers scale based on cluster size.

🛠️ Jobs

A scheduled or triggered pipeline that runs a notebook, JAR, Python file, or Delta Live Table.

• Can have parameters, dependencies, retry logic, alerts
• Supports multi-task workflows
• Runs on job clusters or existing clusters

🌊 Pools

Instance pools reduce cluster startup time and cost by pre-warming VMs.

• Create once → Reuse across multiple jobs
• Ideal for high-frequency jobs
• Saves cost on job cluster creation

Summary Table

# Databricks Notebook: Getting Started with PySpark

# COMMAND ----------
# 📘 1. Read CSV & Parquet from DBFS

# DBFS Path (upload a file via sidebar > Data > Add Data)
csv_path = "/dbfs/tmp/sample.csv"
parquet_path = "/dbfs/tmp/sample.parquet"

# Read CSV
csv_df = spark.read.option("header", True).csv(csv_path)
csv_df.show()

# Write as Parquet
csv_df.write.mode("overwrite").parquet(parquet_path)

# COMMAND ----------
# 📘 2. Connect to Azure Data Lake (ADLS Gen2) securely

# Step 1: Use Azure Key Vault-backed secret scope for credentials
# You must create secret scope via UI or CLI, example:
# dbutils.secrets.get(scope="kv-scope", key="adls-key")

storage_account = "your_storage_account_name"
container = "your_container"
mount_point = "/mnt/adls_mount"

configs = {
  "fs.azure.account.key.%s.dfs.core.windows.net" % storage_account: dbutils.secrets.get(scope="kv-scope", key="adls-key")
}

# Mount ADLS Gen2 (run once per cluster)
dbx_path = "abfss://%s@%s.dfs.core.windows.net/" % (container, storage_account)

try:
  dbutils.fs.mount(source=dbx_path, mount_point=mount_point, extra_configs=configs)
except:
  print("Already mounted")

# List files
display(dbutils.fs.ls(mount_point))

# COMMAND ----------
# 📘 3. Simple PySpark Transformations

# Add column, filter, groupBy
df = csv_df.withColumnRenamed("amount", "sales_amount")
df_filtered = df.filter(df.sales_amount > 100)
df_filtered.groupBy("region").count().show()

# COMMAND ----------
# ✅ END OF HANDS-ON WALKTHROUGH

Here’s the full hands-on Databricks Notebook (First Databricks Notebook) with:

• ✅ Read/write CSV & Parquet from DBFS
• 🔐 Securely connect to ADLS Gen2 via Key Vault
• 🔄 Perform simple PySpark transformations

🎯 Interview Questions & Answers

1. What is Databricks and how is it different from Azure Synapse?

🟡 Databricks excels in big data & AI, Synapse is stronger in SQL + BI integration.

2. Explain the Databricks Workspace structure.

• Workspace: UI to organize assets
  • 📓 Notebooks: Code documents
  • 🛠 Repos: Git-backed version control
  • 🗃 Jobs: Automation pipelines
  • 📁 DBFS: Built-in cloud-backed file store
  • 🔐 Secrets: Secure credential storage

3. What are the components of a Databricks cluster?

• Driver Node: Manages SparkContext, job coordination
• Worker Nodes: Execute Spark tasks
• Libraries: Installed per cluster (PyPI, Maven)
• Runtime: Pre-built Spark environment (e.g., ML, Photon)

📌 A cluster = 1 driver + N workers running Spark jobs.

4. How do you connect ADLS with Databricks securely?

1. Use Azure Key Vault-backed Secret Scope:
   • Store keys/secrets in Azure Key Vault
   • Reference using dbutils.secrets.get(...)
2. Mount ADLS Gen2 using configs: configs = { f"fs.azure.account.key.{account}.dfs.core.windows.net": dbutils.secrets.get(scope="kv-scope", key="adls-key") } dbutils.fs.mount(source="abfss://...", mount_point="/mnt/...", extra_configs=configs)
3. Use ABFS URI or /mnt/ mount in Spark read/write.

Pages: 1 2 3 4 5 6 7 8