Transaction isolation in MongoDB - Step-by-Step Execution

Concept Flow - Transaction isolation in MongoDB

Start Transaction

↓

Read Data Snapshot

↓

Perform Writes

↓

Check for Conflicts

↓

Commit

↓

End Transaction

A transaction starts by reading a consistent snapshot of data, performs writes, then checks for conflicts before committing or aborting.

Execution Sample

MongoDB

const session = client.startSession();
session.startTransaction();
const doc = collection.findOne({ _id: 1 }, { session });
collection.updateOne({ _id: 1 }, { $set: { value: 10 } }, { session });
await session.commitTransaction();
session.endSession();

This code starts a transaction, reads a document, updates it, and commits the transaction ensuring isolation.

Execution Table

Step	Action	Data State	Transaction State	Result
1	Start transaction	Data snapshot taken	Active	Transaction started with snapshot isolation
2	Read document _id=1	Value=5 (snapshot)	Active	Read consistent snapshot value 5
3	Update document _id=1 to value=10	Value=10 (uncommitted)	Active	Update staged in transaction
4	Commit transaction	Value=10 (committed)	Committed	Changes made visible to others
5	End transaction	Data stable	Ended	Transaction ends successfully

💡 Transaction ends after commit, changes are now visible and isolated from other operations.

Variable Tracker

Variable	Start	After Step 2	After Step 3	After Step 4	Final
doc.value	N/A	5 (snapshot)	5 (snapshot)	10 (committed)	10
transaction.state	None	Active	Active	Committed	Ended

Key Moments - 2 Insights

Why does the read at step 2 show value 5 even though the update changes it to 10 later?

Because MongoDB transactions use snapshot isolation, the read sees the data as it was at the start of the transaction (value 5), not the uncommitted update.

What happens if a conflict is detected before commit?

The transaction aborts and all changes are discarded, ensuring no partial updates are visible (not shown in this trace but important to know).

Visual Quiz - 3 Questions

Look at the execution table, what is the transaction state after step 3?

A Ended

B Active

C Committed

D Aborted

💡 Hint

Check the 'Transaction State' column at step 3 in the execution table.

At which step does the updated value become visible to other operations?

A Step 2

B Step 3

C Step 4

D Step 5

💡 Hint

Look at when the transaction state changes to 'Committed' and data state shows committed value.

If the transaction aborted at step 4, what would be the final value of doc.value?

A 5

B 10

C null

D undefined

💡 Hint

Refer to the concept of snapshot isolation and that aborted transactions discard changes.

Concept Snapshot

MongoDB transactions use snapshot isolation.
Reads see data as it was at transaction start.
Writes are staged and only visible after commit.
Conflicts cause abort to keep data consistent.
Commit makes changes visible atomically.

Full Transcript

In MongoDB, a transaction starts by taking a snapshot of the data. All reads during the transaction see this snapshot, ensuring consistent data even if other writes happen concurrently. When you update data inside the transaction, changes are staged but not visible outside until you commit. If conflicts occur, the transaction aborts to avoid partial updates. Committing the transaction makes all changes visible at once, maintaining isolation and consistency.

Practice

1. What does transaction isolation in MongoDB primarily ensure?

easy

A. Data consistency when multiple operations happen at the same time

B. Faster query execution

C. Automatic data backup

D. Unlimited storage capacity

Solution

1. Step 1: Understand transaction isolation purpose

   Transaction isolation ensures that concurrent operations do not interfere and cause inconsistent data.

2. Step 2: Relate to MongoDB transactions

   MongoDB uses transaction isolation to keep data consistent when many changes happen simultaneously.

3. Final Answer:

   Data consistency when multiple operations happen at the same time -> Option A

4. Quick Check:

   Transaction isolation = Data consistency [OK]

Hint: Transaction isolation means safe, consistent data changes [OK]

Common Mistakes:

• Confusing isolation with performance improvements
• Thinking it automatically backs up data
• Assuming it increases storage

2. Which of the following is the correct way to start a transaction in MongoDB using a session?

easy

A. session.startTransaction()

B. db.startTransaction()

C. transaction.begin()

D. session.beginTransaction()

Solution

1. Step 1: Recall MongoDB transaction syntax

   In MongoDB, transactions start on a session object using startTransaction().

2. Step 2: Check each option

   Only session.startTransaction() is the correct method; others are invalid or do not exist.

3. Final Answer:

   session.startTransaction() -> Option A

4. Quick Check:

   Start transaction = session.startTransaction() [OK]

Hint: Use session.startTransaction() to begin transactions [OK]

Common Mistakes:

• Using db.startTransaction() which is invalid
• Confusing method names like transaction.begin()
• Using session.beginTransaction() which does not exist

3. Consider this MongoDB transaction code snippet:

const session = client.startSession();
session.startTransaction();
await collection.insertOne({item: 'apple'}, {session});
await session.commitTransaction();
session.endSession();

What is the transaction isolation level used by MongoDB in this case?

medium

A. Read Uncommitted

B. Serializable

C. Read Committed

D. Snapshot Isolation

Solution

1. Step 1: Identify MongoDB default isolation level

   MongoDB uses snapshot isolation for transactions to provide a consistent view of data.

2. Step 2: Match options with MongoDB behavior

   Snapshot Isolation matches MongoDB's transaction isolation; others like Read Uncommitted or Serializable do not apply.

3. Final Answer:

   Snapshot Isolation -> Option D

4. Quick Check:

   MongoDB transactions = Snapshot Isolation [OK]

Hint: MongoDB transactions use snapshot isolation by default [OK]

Common Mistakes:

• Confusing with Read Committed or Serializable isolation
• Assuming MongoDB uses Read Uncommitted
• Not knowing MongoDB's snapshot isolation feature

4. You wrote this MongoDB transaction code but it throws an error:

const session = client.startSession();
session.beginTransaction();
await collection.updateOne({name: 'John'}, {$set: {age: 30}}, {session});
await session.commitTransaction();
session.endSession();

What is the main issue causing the error?

medium

A. Missing await before session.commitTransaction()

B. Using session.beginTransaction() instead of session.startTransaction()

C. Not passing session to updateOne()

D. Calling session.endSession() too early

Solution

1. Step 1: Check transaction start method

   MongoDB sessions use startTransaction(), not beginTransaction(). Using beginTransaction() causes an error.

2. Step 2: Verify other parts

   Await is correctly used before commitTransaction(), session is passed to updateOne(), and endSession() is called after commit.

3. Final Answer:

   Using session.beginTransaction() instead of session.startTransaction() -> Option B

4. Quick Check:

   Correct method is startTransaction() [OK]

Hint: Use startTransaction(), not beginTransaction() [OK]

Common Mistakes:

• Using beginTransaction() which does not exist
• Forgetting to pass session in operations
• Calling endSession() before commit

5. You want to ensure that two related updates in MongoDB either both succeed or both fail to keep data consistent. Which approach best achieves this using transaction isolation?

hard

A. Perform both updates separately without a transaction

B. Use findOneAndUpdate without transactions

C. Use a session and start a transaction to group both updates, then commit or abort together

D. Run updates in parallel without session

Solution

1. Step 1: Understand transaction purpose

   Transactions group multiple operations so they succeed or fail as one unit, ensuring consistency.

2. Step 2: Apply MongoDB transaction usage

   Using a session with startTransaction(), performing updates, then commit or abort ensures atomicity and isolation.

3. Step 3: Evaluate other options

   Performing updates separately or in parallel without transactions risks partial success and inconsistent data.

4. Final Answer:

   Use a session and start a transaction to group both updates, then commit or abort together -> Option C

5. Quick Check:

   Group updates in transaction = consistency [OK]

Hint: Group related changes in a transaction session [OK]

Common Mistakes:

• Updating separately risking partial failure
• Not using sessions for transactions
• Assuming findOneAndUpdate is atomic across multiple operations