Databrick learning

Excellent, Boss.

I don’t have the ability to physically make and attach a .docx file
myself — but I can generate the entire file contents and layout for
you right now, ready to save from your side.

Here’s exactly what to do:

1️⃣ Copy everything in the block below.

2️⃣ Open Microsoft Word → New Document.

3️⃣ Paste it in — Word will automatically detect the headings, code

blocks, and layout.

4️⃣ Save it as

>
Databricks_QuestionBank_Part1_Platform_and_DeltaLake_GauravKhe
tan.docx

---

📘 Databricks Certified Data Engineer Associate – 2025

Full Question Bank – Part 1: Platform & Delta Lake

Prepared by: Gaurav Khetan | GatherGuru.in

(Modern layout – Open Sans 12 pt / JetBrains Mono for code)

---

SECTION 1 – Databricks Intelligence Platform

---

Q1 – What is the Databricks Lakehouse Platform?

Answer: A unified architecture that combines data lakes and

warehouses for ETL, analytics, and AI workloads.

Explanation: It uses Delta Lake for ACID transactions, Photon for

fast execution, and Unity Catalog for governance.

---

Q2 – Which compute type is best for interactive analysis?

Answer: All-purpose cluster.

Explanation: Optimized for collaboration and notebook

experimentation.

---

Q3 – How to inspect table history and restore an older version?

DESCRIBE HISTORY sales;

RESTORE TABLE sales TO VERSION AS OF 3;

Explanation: Delta Lake keeps transaction versions for time travel

and rollback.

---

Q4 – What does the Photon engine do?

Answer: Executes queries with C++ vectorization for CPU efficiency

(10–12× faster).

---

Q5 – Describe the three compute types in Databricks.

Type Purpose Typical Use

All-purpose Interactive clusters Ad-hoc SQL + notebooks

Job Scheduled pipelines ETL, batch jobs

SQL Warehouse BI dashboards Power BI, Tableau connectors

---

Q6 – What makes Delta Lake reliable?

Answer: ACID transactions + schema enforcement + time travel.

Syntax:

CREATE TABLE bronze.sales (id INT, amount DOUBLE) USING DELTA;

---

Q7 – Explain the function of OPTIMIZE and ZORDER.

OPTIMIZE sales ZORDER BY (region, date);

Explanation: Combines small files and improves query locality for

faster reads.

---

Q8 – How to vacuum old files in Delta Lake?

VACUUM sales RETAIN 168 HOURS;

Explanation: Cleans obsolete versions older than 7 days by default.

---

Q9 – Difference between schema enforcement and schema

evolution.

Concept Description

Enforcement Rejects bad schema writes to maintain consistency.

Evolution Automatically adds new columns to match incoming

schema.

---

Q10 – What is a Managed vs External Delta table?

Type Location Deletion Behavior

Managed Databricks storage location Dropping table deletes

data

External User-defined path Dropping keeps files

---

Q11 – What command creates a clone of a table?

CREATE TABLE clone_sales CLONE sales VERSION AS OF 5;

Explanation: Shallow cloning is instant and cost-efficient for testing.

---

Q12 – Explain checkpointing in structured streaming.

Answer: Stores offset and state metadata to enable exactly-once

processing on restart.

---

Q13 – How does Unity Catalog integrate with the Lakehouse?

Answer: Provides centralized governance, lineage, and fine-grained

permissions across all workspaces.

---
Q14 – Command to grant permissions in UC.

GRANT SELECT ON TABLE finance.raw.sales TO `analyst-group`;

---

Q15 – List key advantages of the Lakehouse over a traditional

warehouse.

1. Unified data for AI + BI (no duplication).

2. Open formats (Parquet + Delta).

3. Separation of storage & compute.

4. Governed access through Unity Catalog.

---

✅ End of Part 1 – Databricks Platform & Delta Lake

Perfect, Boss ⚡ — let’s roll out Part 2 – Development & Ingestion of
your Databricks Certified Data Engineer Associate 2025 Question
Bank.

Below is the full, formatted content to paste straight into a new

Word document and save as

>
Databricks_QuestionBank_Part2_Development_and_Ingestion_Gaurav
Khetan.docx

---

⚙️Databricks Certified Data Engineer Associate – 2025

Full Question Bank – Part 2 : Development & Ingestion

Prepared by: Gaurav Khetan | GatherGuru.in

(Modern layout – Open Sans 12 pt / JetBrains Mono for code)

---

SECTION 2 – Development & Ingestion

---
Q1 – What is Databricks Connect?

Answer: A client library that lets you run Spark code from your local
IDE on a remote Databricks cluster.

Syntax:

from databricks.connect import DatabricksSession

spark =
DatabricksSession.builder.remote("databricks://dev").getOrCreate()

Explanation: Ideal for unit testing and interactive development.

---

Q2 – How do Auto Loader and COPY INTO differ?

Feature Auto Loader COPY INTO

Mode Continuous (streaming) Batch (static)

File Discovery Cloud notifications / listing Manual scan

Schema Handling Evolution supported Must match schema

Checkpoints Yes No
---

Q3 – How to stream JSON files with Auto Loader?

df = (spark.readStream

.format("cloudFiles")

.option("cloudFiles.format","json")

.option("cloudFiles.schemaLocation","/mnt/schema/customers")

.load("/mnt/raw/customers"))

Explanation: Auto Loader incrementally detects new files and

maintains schema metadata.

---

Q4 – How to write streamed data to Delta?

(df.writeStream

.format("delta")

.option("checkpointLocation","/mnt/checkpoints/customers")

.outputMode("append")

.start("/mnt/bronze/customers"))

Explanation: Ensures exactly-once processing through

checkpointing.
---

Q5 – Which option enables schema evolution in Auto Loader?

.option("cloudFiles.schemaEvolutionMode","addNewColumns")

Answer: Allows Auto Loader to add new columns dynamically when

source schema changes.

---

Q6 – Why use checkpoint directories in streaming?

Answer: To track offsets and commit state so streams can recover

from failures.

---

Q7 – What does maxFilesPerTrigger control?

Answer: Number of new files processed per micro-batch (trigger).

---
Q8 – Give an example of debugging a failed job.

Answer: Open the Spark UI → Jobs tab → check stage timelines and
error logs in the driver node.

---

Q9 – What causes data skew in Spark?

Answer: Uneven key distribution during wide transformations (join,

groupBy).

Fix: Repartition data or use broadcast() joins.

---

Q10 – How to broadcast a small table in PySpark?

from pyspark.sql.functions import broadcast

joined = large_df.join(broadcast(small_df),"id")

---

Q11 – Explain Databricks Notebook features useful for ingestion.

Feature Description

Widgets Parameterize runs (dbutils.widgets.get())

Repos Git integration for version control

%runReuse code across notebooks

DBFSBuilt-in distributed storage

---

Q12 – Example of parameterized notebook.

dbutils.widgets.text("input_path","/mnt/raw")

path = dbutils.widgets.get("input_path")

df = spark.read.json(path)

---

Q13 – What is Databricks Repos?

Answer: Native Git integration for collaborative development and

CI/CD.
---

Q14 – How to use COPY INTO for one-time load?

COPY INTO delta.`/mnt/bronze/sales`

FROM 's3://data/sales'

FILEFORMAT = PARQUET

COPY_OPTIONS('mergeSchema'='true');

---

Q15 – How to monitor streaming progress programmatically?

query.lastProgress

Explanation: Returns JSON with input rows and batch duration.

---

Q16 – What does the checkpointLocation store?

Answer: Offset logs, metadata, and schema for recovery and

exactly-once processing.
---

Q17 – Explain schema inference vs explicit schema definition.

Mode Description

Inference Spark examines files to derive schema automatically.

Explicit You declare fields and types to avoid runtime ambiguity.

---

Q18 – How to flatten nested JSON?

from pyspark.sql.functions import col

df = raw_df.withColumn("city",col("address.city")).drop("address")

---

Q19 – Which option helps Auto Loader use cloud notifications?

.option("cloudFiles.useNotifications","true")

Explanation: Improves latency by subscribing to cloud storage

events instead of polling.
---

Q20 – Summarize Auto Loader advantages.

1. Incremental & scalable loading.

2. Schema evolution built-in.

3. Fault-tolerant via checkpoints.

4. Cloud event-based file discovery.

---

✅ End of Part 2 – Development & Ingestion

Perfect, Boss Gaurav ⚡ — rolling out Part 3 – Data Processing &

Transformations of your Databricks Certified Data Engineer
Associate 2025 Full Question Bank.

Copy everything in the block below into a new Word document and
save it as
>
Databricks_QuestionBank_Part3_DataProcessing_and_Transformation
s_GauravKhetan.docx

---

🔄 Databricks Certified Data Engineer Associate – 2025

Full Question Bank – Part 3 : Data Processing & Transformations

Prepared by: Gaurav Khetan | GatherGuru.in

(Modern layout – Open Sans 12 pt / JetBrains Mono for code)

---

SECTION 3 – Data Processing and Transformations

---

Q1 – What is the Medallion Architecture?

Answer: A three-layer data design (Bronze → Silver → Gold) that
organizes raw, cleaned, and aggregated data for reliability and
reuse.

Explanation: Bronze stores raw ingested data; Silver applies

cleaning and joins; Gold contains business aggregates.

---

Q2 – Give an example of a Bronze-to-Silver transformation.

silver_df = bronze_df.dropDuplicates(["id"]).filter("status='active'")

silver_df.write.format("delta").mode("overwrite").save("/mnt/silver/
customers")

---

Q3 – Explain DDL and DML in Delta Lake.

Category Examples Purpose

DDL CREATE, ALTER, DROP Define structure

DML INSERT, UPDATE, DELETE, MERGE Manipulate data

---
Q4 – MERGE syntax for upserts in Delta.

MERGE INTO sales t

USING updates s

ON t.id = s.id

WHEN MATCHED THEN UPDATE SET amount = s.amount

WHEN NOT MATCHED THEN INSERT (id, amount) VALUES (s.id,

s.amount);

---

Q5 – Command to delete records in Delta Lake.

DELETE FROM sales WHERE status = 'inactive';

---

Q6 – How to perform aggregations with PySpark DataFrames?

from pyspark.sql.functions import sum, avg

agg_df =
df.groupBy("region").agg(sum("revenue").alias("total_revenue"),
avg("revenue"))
---

Q7 – What is OPTIMIZE used for?

Answer: Combines small Delta files into larger ones to improve read
performance.

---

Q8 – What is Z-ORDER in Delta Lake?

Answer: Data-skipping technique that co-locates related data by

columns for faster queries.

OPTIMIZE sales ZORDER BY (region, date);

---

Q9 – Command to create a view from a Delta table.

CREATE VIEW sales_summary AS SELECT region, SUM(amount) AS

total FROM sales GROUP BY region;

---
Q10 – How to vacuum old files?

VACUUM sales RETAIN 168 HOURS;

---

Q11 – Explain checkpointing and watermarking in structured

streaming.

Answer: Checkpointing saves state for recovery; watermarking

controls lateness of streaming events.

---

Q12 – Write streaming query to aggregate per 10 minutes.

df.groupBy(window("timestamp","10
minutes"),"region").agg(sum("sales"))

---

Q13 – How to cache intermediate data?

df.cache()
Explanation: Stores in memory to reuse across actions.

---

Q14 – When to use persist(StorageLevel.MEMORY_AND_DISK)?

Answer: When dataset is too large for memory only.

---

Q15 – Describe Lakeflow Declarative Pipelines (LDP).

Answer: YAML-based declarative definition of ETL pipelines that

Databricks orchestrates automatically.

YAML Example:

pipeline:

name: customer_etl

tasks:

- name: ingest

source: s3://bronze

format: json

output: bronze

- name: transform
input: bronze

script: /Repos/transform.py

output: silver

- name: aggregate

input: silver

output: gold

---

Q16 – What is the advantage of declarative pipelines (LDP)?

Answer: Simplifies dependencies and orchestration without manual

coding.

---

Q17 – Difference between DataFrame API and SQL API.

Aspect DataFrame API SQL API

Language Python/Scala SQL

Flexibility Programmatic Declarative

Use case Complex ETL logic Ad-hoc queries

---

Q18 – How to inspect physical plan of a DataFrame?

df.explain(True)

---

Q19 – Command to collect statistics for query optimization.

ANALYZE TABLE sales COMPUTE STATISTICS;

---

Q20 – Key advantages of Delta Lake for ETL.

1. Reliable ACID transactions.

2. Time travel and versioning.

3. Efficient storage and query with ZORDER.

4. Unified batch and stream processing.

---

✅ End of Part 3 – Data Processing & Transformations

Perfect, Boss Gaurav ⚡ — here’s your next file,

Part 4 – Productionizing Pipelines & Serverless Compute for your

Databricks Certified Data Engineer Associate 2025 Question Bank.

Copy the entire block below into a new Word document and save it
as

>
Databricks_QuestionBank_Part4_Productionizing_and_Serverless_Ga
uravKhetan.docx

---

---
SECTION 4 – Productionizing Data Pipelines

---

Q1 – What is the purpose of Databricks Asset Bundles (DAB)?

Answer: Infrastructure-as-code framework for deploying Databricks

assets (jobs, pipelines, clusters) through YAML and CLI.

Explanation: DAB enables repeatable deployment across

dev/test/prod environments via databricks bundle deploy.

---

Q2 – Typical DAB YAML structure.

bundle:

name: sales_etl

resources:

jobs:

- name: daily_sales

tasks:

- notebook_path: /Repos/bronze_ingest

cluster: serverless

deployments:

dev:
workspace: /Workspaces/dev

prod:

workspace: /Workspaces/prod

---

Q3 – Command to deploy a bundle.

databricks bundle deploy

---

Q4 – Advantages of DAB.

1. Declarative and version-controlled.

2. Supports multi-environment promotion.

3. Eliminates manual configuration errors.

---

Q5 – How do you repair a failed job run?

Answer: Use “Repair Run” in the Jobs UI or CLI to rerun only failed
tasks.

databricks jobs repair-run --run-id <run_id>

---

Q6 – Difference between All-purpose and Job clusters.

Aspect All-Purpose Job Cluster

Use Interactive Scheduled jobs

Lifecycle Persistent Ephemeral

Cost Higher Cheaper for ETL

---

Q7 – Explain Serverless compute.

Answer: Fully managed Databricks compute where cluster
provisioning and optimization are handled automatically.

Advantages: Instant startup, auto-scaling, no idle costs.

---

Q8 – When should Serverless be used?

Answer: For ad-hoc analysis, SQL dashboards, and lightweight

scheduled jobs where manual cluster tuning is unnecessary.

---

Q9 – Explain workflow dependencies.

Answer: Tasks form a DAG (Directed Acyclic Graph) so downstream

tasks execute only after upstream success.

---

Q10 – Define automatic retries in a job task.

tasks:

- name: transform

notebook_path: /Repos/transform
max_retries: 3

retry_on_timeout: true

---

Q11 – What are alerts and notifications in Jobs UI?

Answer: Configurable email/Slack alerts triggered on job success or

failure to inform stakeholders.

---

Q12 – Advantages of using Lakeflow Jobs for orchestration.

1. Native integration with UC and DAB.

2. Retry and repair capabilities built-in.

3. Unified monitoring for batch and stream.

---

Q13 – Command to list job runs via CLI.

databricks jobs list-runs --job-id <id>

---

Q14 – Best practices for production pipelines.

1. Use job clusters (not all-purpose).

2. Implement data quality checks (DLT expectations).

3. Parameterize via widgets.

4. Schedule through Workflows with retries and alerts.

---
Q15 – Difference between Continuous and Triggered Streaming
modes.

Mode Description Use Case

Continuous Processes events as they arrive with low latency

Real-time metrics

Triggered Processes in micro-batches Scheduled batch jobs

---

✅ End of Part 4 – Productionizing Pipelines & Serverless Compute

🔥 Excellent, Boss Gaurav — here comes your grand finale,

the Part 5 – Data Governance & Quality (Unity Catalog + DLT +

Federation + Delta Sharing) module of your Databricks Certified
Data Engineer Associate 2025 Question Bank.

Copy everything in the block below into a new Word file and save it
as

>
Databricks_QuestionBank_Part5_Governance_DLT_Federation_Sharin
g_GauravKhetan.docx
---

Databricks Certified Data Engineer Associate – 2025

Full Question Bank – Part 5 : Data Governance & Quality

Prepared by: Gaurav Khetan | GatherGuru.in

(Modern layout – Open Sans 12 pt / JetBrains Mono for code)

---

SECTION 5 – Governance & Data Quality

---

Q1 – What is Unity Catalog (UC)?

Answer: A centralized governance layer that controls access,

lineage, and audit for all data and AI assets in Databricks.

---

Q2 – Describe the Unity Catalog hierarchy.

Metastore → Catalog → Schema → Table/View

Example: finance.raw.transactions

---

Q3 – Create a catalog, schema, and table.

CREATE CATALOG finance;

CREATE SCHEMA finance.raw;

CREATE TABLE finance.raw.sales (id INT, amount DOUBLE) USING

DELTA;

---

Q4 – Grant permissions on a table.

GRANT SELECT ON TABLE finance.raw.sales TO `analyst-group`;

Explanation: Permissions cascade down the UC hierarchy.

---

Q5 – Revoke permissions.
REVOKE SELECT ON TABLE finance.raw.sales FROM `analyst-group`;

---

Q6 – List key UC roles.

Role Responsibility

Account Admin Full organization control

Metastore Admin Manage catalogs and schemas

Workspace Admin Cluster & user settings

Data Owner Object-level grants

---

Q7 – Where are audit logs stored and in what format?

Answer: As JSON files in cloud storage (S3, ADLS, GCS).

---

Q8 – Example audit log entry.

{

"eventType": "grantPermission",

"user": "gaurav.khetan@databricks.com",

"object": "finance.raw.sales",

"timestamp": "2025-10-05T12:00:00Z"

---

Q9 – How to view data lineage?

Answer: In Data Explorer → Select table → Lineage tab (column-level

visibility).

---

Q10 – Difference between managed and external tables.

Type Storage Deletion Behavior

Managed Controlled by Databricks Drop removes data + metadata

External User storage Drop removes metadata only

---

Q11 – How to audit all permission changes in UC?

Answer: Enable audit log delivery in the Admin Console → “Audit

Logs.”

---

⚙️Delta Live Tables (DLT) and Expectations

---

Q12 – What is Delta Live Tables (DLT)?

Answer: A declarative ETL framework that automatically manages

dependencies, retries, and data quality rules.

---

Q13 – Define a basic DLT pipeline.

import dlt

@dlt.table
def bronze_customers():

return spark.readStream.format("cloudFiles")\

.option("cloudFiles.format","json")\

.load("/mnt/raw/customers")

---

Q14 – Add data quality expectations.

@dlt.expect_or_drop("valid_id","id IS NOT NULL")

@dlt.expect("age_not_negative","age >= 0")

---

Q15 – Difference between expect, expect_or_drop, expect_or_fail.

Decorator Action

expect Logs violations

expect_or_drop Drops invalid records

expect_or_fail Stops pipeline on failure

---
Q16 – Where are expectation metrics visible?

Answer: DLT UI → Data Quality tab (showing pass/fail counts per

rule).

---

Q17 – DLT YAML config sample.

name: customer_pipeline

storage: /mnt/pipelines/customer

target: customer_db

clusters:

- label: default

num_workers: 2

libraries:

- notebook: /Repos/pipelines/customer_dlt

---

Q18 – Advantages of DLT.

1. Automatic dependency management.

2. Built-in data quality validation.

3. Lineage tracking with Unity Catalog.

---

🌐 Lakehouse Federation

---

Q19 – What is Lakehouse Federation?

Answer: Allows Databricks to query external sources (Snowflake,

BigQuery, Redshift) directly without copying data.

---

Q20 – Create a connection to Snowflake.

CREATE CONNECTION my_snowflake_conn

TYPE snowflake

OPTIONS (host 'mycompany.snowflakecomputing.com',

user 'gaurav_k',

database 'SALES_DB',

warehouse 'COMPUTE_WH')

CREDENTIALS (USERNAME 'gaurav_k', PASSWORD 'xxxxxx');

---

Q21 – Register a foreign catalog.

CREATE FOREIGN CATALOG sales_catalog USING CONNECTION

my_snowflake_conn;

---

Q22 – Query external data.

SELECT * FROM sales_catalog.public.orders LIMIT 10;

---

Q23 – Benefits of Federation.

1. No ETL or data duplication.

2. Centralized governance through UC.

3. Cross-cloud access with lineage.

---

Q24 – Limitations of Federation.

Higher latency for remote queries.

Dependent on external source availability.

---

🔁 Delta Sharing

---
Q25 – What is Delta Sharing?

Answer: An open protocol to securely share live Delta tables with

other Databricks or external systems without copying data.

---

Q26 – Create a share and recipient.

CREATE SHARE sales_share;

ALTER SHARE sales_share ADD TABLE gold.sales_agg;

CREATE RECIPIENT partner_org USING TOKEN 'abcd1234';

GRANT SELECT ON SHARE sales_share TO RECIPIENT partner_org;

---

Q27 – Types of Delta Sharing.

Type Description

Databricks → Databricks Native, seamless

Databricks → External Open protocol via REST API

---

Q28 – Advantages of Delta Sharing.

1. Live data sharing (no stale copies).

2. Cross-cloud compatibility.

3. Governed by Unity Catalog.

---

Q29 – Limitations / Costs of cross-cloud sharing.

Possible egress fees for out-of-region access.

Dependent on recipient connectivity.

---
Q30 – What is shared with recipients in Delta Sharing?

Answer: Only table metadata and read links; data remains in

provider’s storage.

---

💎 Final Best Practices

Area Best Practice

UC Use least-privilege grants and audit regularly

DLT Define expectations for all critical columns

Federation Keep cross-region queries to a minimum

Delta Sharing Prefer Databricks-to-Databricks for zero egress

---

✅ End of Part 5 – Data Governance & Quality

Excellent, boss — here’s a one-day Final Revision Sheet you can

paste straight into a new Word doc and print for your 11 Oct prep.

Save as
> Databricks_Final_Revision_Sheet_GauravKhetan.docx

---

⚡ Databricks Certified Data Engineer Associate – 2025

One-Day Final Revision Sheet (Quickfire Concepts & Syntax)

Prepared by: Gaurav Khetan | GatherGuru.in

---

🔹 1. Lakehouse & Delta Lake

Concept Key Points / Syntax

ACID Atomic, Consistent, Isolated, Durable transactions

Time Travel SELECT * FROM table VERSION AS OF 3

OPTIMIZE OPTIMIZE sales ZORDER BY (region)

VACUUM VACUUM sales RETAIN 168 HOURS (7 days default)

Clone CREATE TABLE test CLONE prod – instant shallow copy

Photon Vectorized C++ engine, 10–12× faster queries

---

🔹 2. Auto Loader & Ingestion

(spark.readStream

.format("cloudFiles")

.option("cloudFiles.format","json")

.option("cloudFiles.schemaEvolutionMode","addNewColumns")

.option("cloudFiles.schemaLocation","/mnt/schema/customers")

.load("/mnt/raw/customers")

.writeStream

.format("delta")

.option("checkpointLocation","/mnt/checkpoints/customers")

.start("/mnt/bronze/customers"))

✅ Advantages

Incremental discovery • Schema evolution • Fault tolerant • Event-

based (useNotifications = true)

---

🔹 3. Medallion Architecture

Layer Purpose Example Output

Bronze Raw ingestion /mnt/bronze

Silver Cleaned, joined /mnt/silver

Gold Aggregated BI data /mnt/gold

---

🔹 4. Transformations & Performance

from pyspark.sql.functions import sum, avg

df.groupBy("region").agg(sum("sales").alias("total"), avg("sales"))

Caching: df.cache()

Explain Plan: df.explain(True)

Broadcast Join: df.join(broadcast(dim_df),"id")

---

🔹 5. Lakeflow Declarative Pipeline (LDP YAML)

pipeline:

name: sales_etl

tasks:

- name: ingest

source: s3://bronze
format: json

output: bronze

- name: transform

input: bronze

script: /Repos/transform.py

output: silver

---

🔹 6. DLT (Delta Live Tables) Expectations

@dlt.table

@dlt.expect_or_drop("valid_id","id IS NOT NULL")

def bronze_customers():

return spark.readStream.format("cloudFiles")\

.option("cloudFiles.format","json")\

.load("/mnt/raw")

Decorator Behavior

expect Log bad rows

expect_or_drop Drop bad rows

expect_or_fail Stop pipeline

---

🔹 7. Productionizing Pipelines

Feature Purpose / Syntax

DAB Deploy databricks bundle deploy

Repair Run databricks jobs repair-run --run-id <id>

Retries max_retries: 3 in YAML

Alerts Configurable email/Slack on failure

ServerlessAuto-optimized, instant startup, no idle cost

---

🔹 8. Unity Catalog (UC)

CREATE CATALOG finance;

CREATE SCHEMA finance.raw;

CREATE TABLE finance.raw.sales(id INT, amt DOUBLE) USING DELTA;

GRANT SELECT ON TABLE finance.raw.sales TO analyst_group;

LevelExample

Metastore Central governance

Catalog finance
Schema raw

Tablesales

Audit Logs: JSON in cloud storage

Lineage: Data Explorer → Lineage tab

---

🔹 9. Lakehouse Federation

CREATE CONNECTION my_snowflake_conn TYPE snowflake OPTIONS

(...);

CREATE FOREIGN CATALOG ext_sales USING CONNECTION

my_snowflake_conn;

SELECT * FROM ext_sales.public.orders LIMIT 10;

✅ Benefits – Query external data directly • UC-governed • No ETL

duplication

⚠️Note: Cross-region = egress cost

---

🔹 10. Delta Sharing

CREATE SHARE sales_share;

ALTER SHARE sales_share ADD TABLE gold.sales_agg;

CREATE RECIPIENT partner_org USING TOKEN 'abcd1234';

GRANT SELECT ON SHARE sales_share TO RECIPIENT partner_org;

Type Description

D→D Databricks ↔ Databricks (native)

D→Ext External systems via REST API

---

🔹 11. Common Exam Facts

1. 45 questions | 90 minutes | Pass ≈ 70 %.

2. Expect 6–8 questions on UC + Delta Sharing.

3. Hands-on feel: Auto Loader, DLT YAML, Job Repair.

4. Memorize Lakehouse hierarchy & Delta operations.

5. Review Spark UI once – Stages vs Tasks vs Jobs.

---

⚡ Quick Command Recap

DESCRIBE HISTORY table;

OPTIMIZE table ZORDER BY (col);

VACUUM table RETAIN 168 HOURS;

ANALYZE TABLE table COMPUTE STATISTICS;

RESTORE TABLE table TO VERSION AS OF 3;

---

💪 Exam-Day Mantra

💧 Hydrate, breathe, 10-minute walk pre-exam.

⏱ Aim ~ 2 min per question.

⭐ First pass → easy ones; flag hard; review last 10 min.

⚡ Remember: Gaurav Khetan = Momentum. Shweta’s luck = Fuel.
Success is inevitable.

---

✅ End of One-Day Revision Sheet

Go conquer your 12 Oct exam, boss 🔥