Batching and linger configuration in Kafka - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is the purpose of batching in Kafka producers?

Batching groups multiple records into a single request to improve throughput and reduce network overhead.

beginner

What does the 'linger.ms' configuration control in Kafka producers?

It controls how long the producer waits before sending a batch, allowing more records to accumulate for efficient batching.

intermediate

How does increasing 'batch.size' affect Kafka producer performance?

Increasing 'batch.size' allows larger batches, which can improve throughput but may increase latency if batches take longer to fill.

intermediate

What happens if 'linger.ms' is set to 0 in Kafka producer configuration?

The producer sends records immediately without waiting to batch, which can reduce latency but may increase network overhead.

advanced

Explain the trade-off between latency and throughput when tuning 'batch.size' and 'linger.ms'.

Larger batch sizes and longer linger times increase throughput by sending more data at once but add latency because records wait longer before sending.

What does the 'batch.size' setting in Kafka producer specify?

A Maximum number of batches per second

B Maximum size of a batch in bytes

C Time to wait before sending a batch

D Number of retries for sending a batch

'batch.size' sets the maximum size in bytes for a batch of records sent by the producer.

If you want to reduce latency in Kafka producer, what should you do with 'linger.ms'?

A Disable batching

B Set it to a higher value

C Increase 'batch.size'

D Set it to zero

Setting 'linger.ms' to zero makes the producer send records immediately, reducing latency.

What is a potential downside of setting a very large 'batch.size'?

A More frequent requests to broker

B Increased network overhead

C Higher latency due to waiting for batch to fill

D Lower throughput

A large batch size can increase latency because the producer waits longer to fill the batch before sending.

Which configuration helps Kafka producer to wait for more records before sending?

A linger.ms

B retries

C acks

D batch.size

'linger.ms' controls how long the producer waits to accumulate more records before sending a batch.

What is the main benefit of batching in Kafka producers?

A Reduces network overhead and improves throughput

B Guarantees message delivery

C Improves message ordering

D Encrypts messages

Batching reduces network overhead by sending multiple records together, improving throughput.

Describe how 'batch.size' and 'linger.ms' work together in Kafka producer batching.

Think about size and time as two ways to decide when to send data.

Explain the trade-offs when increasing 'linger.ms' in Kafka producer configuration.

Consider what happens when you wait longer to send data.

Practice

1. What does the linger.ms configuration control in Kafka producer settings?

easy

A. The maximum size of a message batch in bytes

B. The maximum time to wait before sending a batch of messages

C. The number of retries for sending a message

D. The timeout for consumer group rebalancing

Solution

1. Step 1: Understand the role of linger.ms

   The linger.ms setting defines how long the producer waits before sending a batch, allowing more messages to accumulate.

2. Step 2: Differentiate from other settings

   Other options like batch size control bytes, retries control attempts, and consumer timeout are unrelated to linger time.

3. Final Answer:

   The maximum time to wait before sending a batch of messages -> Option B

4. Quick Check:

   linger.ms = wait time before batch send [OK]

Hint: linger.ms sets wait time before sending batch [OK]

Common Mistakes:

• Confusing linger.ms with batch size
• Thinking linger.ms controls retries
• Mixing producer and consumer settings

2. Which of the following is the correct way to set the batch size to 32 KB in Kafka producer configuration?

easy

A. "batch.size": 32768

B. "batch.size": "32KB"

C. "batch.size": 32000

D. "batch.size": "32kb"

Solution

1. Step 1: Understand batch.size unit

   The batch.size is specified in bytes, so 32 KB equals 32768 bytes.

2. Step 2: Check option formats

   Options with strings like "32KB" or "32kb" are invalid because batch.size expects an integer number of bytes.

3. Final Answer:

   "batch.size": 32768 -> Option A

4. Quick Check:

   batch.size = bytes number, 32 KB = 32768 [OK]

Hint: batch.size must be integer bytes, not string [OK]

Common Mistakes:

• Using string values instead of integers
• Confusing KB with bytes count
• Using lowercase 'kb' which is invalid

3. Given the following Kafka producer settings:

{
  "batch.size": 16384,
  "linger.ms": 100
}

If messages arrive every 10 ms, what is the expected behavior?

medium

A. Messages are sent immediately without batching

B. Messages are delayed indefinitely until batch size is reached

C. Messages are sent only when batch size reaches 16384 bytes, ignoring linger.ms

D. Messages are batched and sent every 100 ms or when batch size is reached

Solution

1. Step 1: Analyze linger.ms and batch.size interaction

   The producer waits up to 100 ms (linger.ms) to accumulate messages or sends earlier if batch size (16384 bytes) is reached.

2. Step 2: Consider message arrival rate

   Messages arrive every 10 ms, so batching will likely happen before 100 ms or batch size limit, whichever comes first.

3. Final Answer:

   Messages are batched and sent every 100 ms or when batch size is reached -> Option D

4. Quick Check:

   linger.ms = max wait, batch.size = max batch [OK]

Hint: linger.ms sets max wait, batch.size sets max batch [OK]

Common Mistakes:

• Ignoring linger.ms and assuming immediate send
• Thinking linger.ms is minimum wait time
• Assuming messages delay indefinitely

4. You configured Kafka producer with:

{
  "batch.size": 0,
  "linger.ms": 500
}

What is the problem with this configuration?

medium

A. batch.size cannot be zero; it disables batching causing errors

B. linger.ms must be zero or less to disable delay

C. batch.size zero means messages are sent immediately without batching

D. No problem; this configuration is valid and efficient

Solution

1. Step 1: Understand batch.size = 0 effect

   Setting batch.size to zero disables batching, so each message is sent immediately.

2. Step 2: Check linger.ms impact

   Even with linger.ms = 500 ms, messages won't batch because batch.size = 0 forces immediate send.

3. Final Answer:

   batch.size zero means messages are sent immediately without batching -> Option C

4. Quick Check:

   batch.size=0 disables batching, linger.ms ignored [OK]

Hint: batch.size=0 disables batching, sends immediately [OK]

Common Mistakes:

• Assuming batch.size=0 causes errors
• Thinking linger.ms overrides batch.size=0
• Believing this config improves efficiency

5. You want to optimize Kafka producer for low latency but still batch messages efficiently. Which configuration is best?

hard

A. "batch.size": 65536, "linger.ms": 5

B. "batch.size": 16384, "linger.ms": 500

C. "batch.size": 0, "linger.ms": 0

D. "batch.size": 32768, "linger.ms": 1000

Solution

1. Step 1: Define low latency and efficient batching

   Low latency means sending messages quickly; efficient batching means grouping enough messages to reduce overhead.

2. Step 2: Evaluate options

   "batch.size": 65536, "linger.ms": 5 uses a large batch size (65536 bytes) and a very low linger.ms (5 ms), allowing quick sends with good batching. Other options have too high linger.ms or batch.size=0 disables batching.

3. Final Answer:

   "batch.size": 65536, "linger.ms": 5 -> Option A

4. Quick Check:

   Low linger.ms + large batch = low latency + efficient batching [OK]

Hint: Low linger.ms + large batch size = fast and efficient [OK]

Common Mistakes:

• Choosing linger.ms too high for low latency
• Setting batch.size to zero disables batching
• Ignoring balance between batch size and linger.ms