Auto-commit vs manual commit in Kafka - Quick Revision & Key Differences

Recall & Review

beginner

What is auto-commit in Kafka consumer?

Auto-commit is when Kafka automatically saves the position of the last message read (offset) at regular intervals without programmer intervention.

beginner

What does manual commit mean in Kafka consumer?

Manual commit means the programmer explicitly tells Kafka when to save the offset after processing messages, giving more control over message handling.

intermediate

Why might you prefer manual commit over auto-commit?

Manual commit is preferred when you want to ensure messages are fully processed before saving the offset, avoiding message loss or duplication.

intermediate

What is a risk of using auto-commit in Kafka?

Auto-commit can cause message loss or duplication if the consumer crashes after committing but before processing messages fully.

beginner

How does Kafka track which messages a consumer has read?

Kafka tracks this using offsets, which are numbers marking the position of the last message read in a topic partition.

What does auto-commit do in Kafka consumers?

A Requires manual saving of offsets

B Automatically saves the offset at intervals

C Deletes messages after reading

D Sends messages to another topic

Auto-commit automatically saves the offset at regular intervals without manual intervention.

Why use manual commit in Kafka?

A To avoid using offsets

B To speed up message consumption

C To automatically delete messages

D To control when offsets are saved after processing

Manual commit lets you decide when to save offsets, ensuring messages are processed before committing.

What risk is associated with auto-commit?

A Offsets never saved

B Slower message processing

C Message loss or duplication if crash happens after commit

D Messages get corrupted

If a consumer crashes after auto-commit but before processing, messages may be lost or duplicated.

Which Kafka feature tracks the last read message?

A Offset

B Partition

C Topic

D Broker

Offsets mark the position of the last message read in a partition.

If you want to ensure no message is lost, which commit method is better?

A Manual commit

B Auto-commit

C No commit

D Random commit

Manual commit allows you to commit offsets only after successful processing, reducing message loss.

Explain the difference between auto-commit and manual commit in Kafka consumers.

Think about who controls when the offset is saved.

Why might manual commit be safer than auto-commit when processing messages?

Consider what happens if the consumer crashes.

Practice

1. What is the main difference between auto-commit and manual commit in Kafka consumer offset management?

easy

A. Auto-commit requires manual calls to commit offsets, manual commit commits automatically.

B. Auto-commit disables offset tracking, manual commit enables it.

C. Auto-commit commits offsets automatically at intervals, manual commit requires explicit commit calls.

D. Auto-commit commits offsets only after consumer shutdown, manual commit commits continuously.

Solution

1. Step 1: Understand auto-commit behavior

   Auto-commit commits offsets automatically at regular intervals without user intervention.

2. Step 2: Understand manual commit behavior

   Manual commit requires the consumer to explicitly call commit() after processing messages.

3. Final Answer:

   Auto-commit commits offsets automatically at intervals, manual commit requires explicit commit calls. -> Option C

4. Quick Check:

   Auto-commit = automatic, Manual commit = explicit call [OK]

Hint: Auto-commit is automatic; manual commit needs your call [OK]

Common Mistakes:

• Confusing which method commits automatically
• Thinking manual commit disables offset tracking
• Believing auto-commit commits only on shutdown

2. Which of the following is the correct way to disable auto-commit in Kafka consumer configuration?

easy

A. "enable.auto.commit" : false

B. "auto.commit.enable" : false

C. "auto.commit" : "disable"

D. "enable.auto.commit" : true

Solution

1. Step 1: Identify the correct config key

   The correct Kafka consumer config key to control auto-commit is "enable.auto.commit".

2. Step 2: Identify the correct value to disable auto-commit

   Setting "enable.auto.commit" to false disables auto-commit.

3. Final Answer:

   "enable.auto.commit" : false -> Option A

4. Quick Check:

   Disable auto-commit = enable.auto.commit = false [OK]

Hint: Use "enable.auto.commit" = false to disable auto-commit [OK]

Common Mistakes:

• Using wrong config key names
• Setting true instead of false to disable
• Using string values instead of boolean

3. Given this Kafka consumer code snippet with manual commit:

consumer.poll(Duration.ofMillis(100)).forEach(record -> {
    process(record);
});
consumer.commitSync();

What happens if commitSync() is called before processing the records?

medium

A. Offsets are committed before processing, risking message loss on failure.

B. Offsets are committed after processing, ensuring safe processing.

C. commitSync() throws an exception if called before processing.

D. Offsets are not committed at all.

Solution

1. Step 1: Understand commitSync timing

   commitSync() commits the current offsets to Kafka immediately when called.

2. Step 2: Analyze calling commitSync before processing

   If commitSync() is called before processing, offsets are saved before messages are handled, risking data loss if processing fails.

3. Final Answer:

   Offsets are committed before processing, risking message loss on failure. -> Option A

4. Quick Check:

   Commit before processing = risk of message loss [OK]

Hint: Commit offsets only after processing messages [OK]

Common Mistakes:

• Assuming commitSync waits for processing
• Thinking commitSync throws error if called early
• Believing offsets commit automatically without call

4. You have disabled auto-commit and use manual commit in your Kafka consumer. You notice some offsets are not committed, causing message reprocessing. What is the likely cause?

medium

A. Kafka broker is down, so commits fail silently.

B. Auto-commit is enabled, causing conflicts.

C. You called commit() before processing messages.

D. You forgot to call commit() after processing messages.

Solution

1. Step 1: Check manual commit requirements

   When auto-commit is disabled, the consumer must call commit() explicitly after processing.

2. Step 2: Identify cause of missing commits

   If offsets are not committed, likely commit() was not called after processing, causing reprocessing.

3. Final Answer:

   You forgot to call commit() after processing messages. -> Option D

4. Quick Check:

   Manual commit needs explicit commit() call [OK]

Hint: Always call commit() after processing when auto-commit is off [OK]

Common Mistakes:

• Assuming auto-commit still works when disabled
• Blaming Kafka broker without checking code
• Calling commit() before processing messages

5. You want to ensure exactly-once processing in a Kafka consumer application. Which commit strategy is safer and why?

hard

A. Auto-commit, because it commits offsets automatically without extra code.

B. Manual commit, because you commit offsets only after successful processing.

C. Auto-commit, because it commits offsets after processing each message.

D. Manual commit, because it commits offsets before processing messages.

Solution

1. Step 1: Understand exactly-once processing needs

   Exactly-once means processing each message once and committing offset only after success.

2. Step 2: Compare commit strategies

   Auto-commit commits offsets automatically at intervals, risking committing before processing completes. Manual commit lets you commit only after successful processing.

3. Final Answer:

   Manual commit, because you commit offsets only after successful processing. -> Option B

4. Quick Check:

   Exactly-once needs manual commit after success [OK]

Hint: Commit manually after success for exactly-once processing [OK]

Common Mistakes:

• Assuming auto-commit guarantees exactly-once
• Committing offsets before processing
• Ignoring commit timing importance