Read concern and write concern in transactions in MongoDB

Introduction

Read concern and write concern help control how data is read and saved safely during transactions.

When you want to make sure you read the most recent data in a transaction.

When you want to ensure your data changes are saved to multiple servers before finishing.

When you want to avoid reading data that might be rolled back later.

When you want to balance speed and safety in your database operations.

When you want to control how strict the database is about data consistency.

Syntax

MongoDB

session.startTransaction({ readConcern: { level: "level" }, writeConcern: { w: number, wtimeout: milliseconds } })

readConcern levels include: 'local', 'majority', 'linearizable', 'snapshot'.

writeConcern 'w' specifies how many servers must confirm the write.

Examples

Starts a transaction that reads data confirmed by most servers and waits up to 5 seconds for writes to be saved by most servers.

MongoDB

session.startTransaction({ readConcern: { level: "majority" }, writeConcern: { w: "majority", wtimeout: 5000 } })

Starts a transaction that reads the latest data on the current server and requires only one server to confirm writes.

MongoDB

session.startTransaction({ readConcern: { level: "local" }, writeConcern: { w: 1 } })

Sample Program

This code starts a transaction with majority read and write concerns, inserts a document, commits the transaction, ends the session, and then finds the inserted document.

MongoDB

const session = db.getMongo().startSession();
session.startTransaction({ readConcern: { level: "majority" }, writeConcern: { w: "majority", wtimeout: 1000 } });
const coll = session.getDatabase("test").myCollection;
coll.insertOne({ item: "apple", qty: 10 });
await session.commitTransaction();
session.endSession();
db.myCollection.find({ item: "apple" }).toArray();

Important Notes

Using 'majority' readConcern ensures you read data confirmed by most servers, avoiding reading uncommitted changes.

writeConcern with 'w: majority' waits for most servers to confirm the write, increasing data safety.

Higher read and write concerns can slow down operations but improve data consistency.

Summary

Read concern controls how fresh or confirmed the data you read is during a transaction.

Write concern controls how many servers must confirm your data changes before the transaction commits.

Choosing the right concerns balances speed and data safety.

Practice

1. What does readConcern control in a MongoDB transaction?

easy

A. How fresh or confirmed the data you read is during the transaction

B. How many servers confirm your data changes before commit

C. The speed of the transaction execution

D. The size of the data being written

Solution

1. Step 1: Understand the role of readConcern

   Read concern defines the level of isolation and freshness of data read during a transaction.

2. Step 2: Differentiate from writeConcern

   Write concern controls confirmation of writes, not reads.

3. Final Answer:

   How fresh or confirmed the data you read is during the transaction -> Option A

4. Quick Check:

   readConcern = data freshness [OK]

Hint: Read concern is about data freshness, write concern is about confirmation [OK]

Common Mistakes:

• Confusing readConcern with writeConcern
• Thinking readConcern controls write confirmations
• Assuming readConcern affects transaction speed

2. Which of the following is the correct way to specify writeConcern in a MongoDB transaction?

easy

A. { writeConcern: { w: 'majority' } }

B. { readConcern: { w: 'majority' } }

C. { writeConcern: { r: 'majority' } }

D. { writeConcern: 'readConcern' }

Solution

1. Step 1: Recall writeConcern syntax

   Write concern uses the key w to specify acknowledgement level, e.g., 'majority'.

2. Step 2: Check options for correctness

   { writeConcern: { w: 'majority' } } correctly uses writeConcern: { w: 'majority' }. Others misuse keys or mix readConcern and writeConcern.

3. Final Answer:

   { writeConcern: { w: 'majority' } } -> Option A

4. Quick Check:

   writeConcern uses w key [OK]

Hint: writeConcern uses w, readConcern uses level [OK]

Common Mistakes:

• Using 'r' instead of 'w' for writeConcern
• Mixing readConcern and writeConcern keys
• Passing writeConcern as a string instead of object

3. Given this transaction code snippet, what is the effective read concern level?

const session = client.startSession();
session.startTransaction({ readConcern: { level: 'snapshot' }, writeConcern: { w: 'majority' } });
// transaction operations
await session.commitTransaction();

medium

A. local

B. majority

C. linearizable

D. snapshot

Solution

1. Step 1: Identify readConcern in transaction options

   The transaction is started with readConcern: { level: 'snapshot' }.

2. Step 2: Understand readConcern levels

   'snapshot' means the transaction reads a consistent snapshot of data at start time.

3. Final Answer:

   snapshot -> Option D

4. Quick Check:

   readConcern level = snapshot [OK]

Hint: Check readConcern.level in startTransaction options [OK]

Common Mistakes:

• Confusing writeConcern level with readConcern level
• Assuming default readConcern instead of specified
• Mixing up 'majority' with readConcern levels

4. Identify the error in this transaction code snippet:

const session = client.startSession();
session.startTransaction({ readConcern: { level: 'majority' }, writeConcern: { w: 1 } });
// transaction operations
await session.commitTransaction();

medium

A. writeConcern w: 1 is invalid

B. Missing session.endSession() call

C. readConcern level 'majority' is invalid

D. startTransaction syntax is incorrect

Solution

1. Step 1: Check for session lifecycle management

   The code does not call session.endSession() after commitTransaction(). This is necessary to properly release session resources and avoid leaks.

2. Step 2: Verify other components

   readConcern level 'majority' is valid; writeConcern { w: 1 } is valid (acknowledgement from primary); syntax is correct.

3. Final Answer:

   Missing session.endSession() call -> Option B

4. Quick Check:

   endSession after commit/abort [OK]

Hint: Always call session.endSession() after commitTransaction or abortTransaction [OK]

Common Mistakes:

• Thinking readConcern 'majority' is invalid for transactions
• Assuming writeConcern w:1 is invalid
• Believing session.endSession() is optional

5. You want a transaction that reads the most recent confirmed data and ensures all writes are acknowledged by a majority of servers. Which combination of readConcern and writeConcern should you use?

hard

A. readConcern: 'snapshot', writeConcern: { w: 0 }

B. readConcern: 'local', writeConcern: { w: 1 }

C. readConcern: 'majority', writeConcern: { w: 'majority' }

D. readConcern: 'linearizable', writeConcern: { w: 'majority' }

Solution

1. Step 1: Understand readConcern 'majority'

   This level reads data confirmed by a majority of servers, ensuring recent confirmed data.

2. Step 2: Understand writeConcern { w: 'majority' }

   This ensures writes are acknowledged by a majority of servers before commit.

3. Step 3: Evaluate other options

   readConcern: 'local', writeConcern: { w: 1 } uses weaker writeConcern; C uses no write acknowledgement; D uses linearizable readConcern which is stricter than needed.

4. Final Answer:

   readConcern: 'majority', writeConcern: { w: 'majority' } -> Option C

5. Quick Check:

   Majority read and write concern = readConcern: 'majority', writeConcern: { w: 'majority' } [OK]

Hint: Use 'majority' for both read and write for confirmed data [OK]

Common Mistakes:

• Choosing writeConcern w:1 for majority confirmation
• Confusing 'snapshot' with 'majority' readConcern
• Ignoring writeConcern w:0 means no acknowledgement