Client metrics monitoring in Kafka - Step-by-Step Execution

Process Flow - Client metrics monitoring

Start Kafka Client

↓

Initialize Metrics Collection

↓

Collect Metrics (e.g., requests, errors, latency)

↓

Send Metrics to Monitoring System

↓

Analyze Metrics & Alert if needed

↓

Repeat Collection Periodically

↩Back to Collect Metrics

The Kafka client starts and initializes metrics collection, then periodically collects and sends metrics to a monitoring system for analysis and alerting.

Execution Sample

Kafka

props.put("metric.reporters", "org.apache.kafka.common.metrics.JmxReporter");
KafkaConsumer<String, String> consumer = new KafkaConsumer<>(props);
consumer.subscribe(List.of("topic1"));
// Metrics collected automatically
// Access metrics via JMX or custom reporter

This code configures a Kafka consumer to report metrics via JMX, subscribes to a topic, and enables automatic metrics collection.

Process Table

Step	Action	Metrics Collected	Metrics Sent	Monitoring System Response
1	Kafka client starts and initializes metrics	None yet	None	Waiting
2	Consumer subscribes to topic	Subscription count=1	Subscription metric sent	Acknowledged
3	Consume messages batch 1	Messages consumed=100, Errors=0, Latency=20ms	Metrics sent	Metrics received
4	Consume messages batch 2	Messages consumed=120, Errors=1, Latency=25ms	Metrics sent	Alert triggered for error
5	Consume messages batch 3	Messages consumed=110, Errors=0, Latency=22ms	Metrics sent	Metrics received
6	Periodic metrics collection continues	Updated metrics	Metrics sent	Monitoring ongoing
7	Stop client	Final metrics collected	Final metrics sent	Monitoring stopped

💡 Client stopped, metrics collection and reporting ended

Status Tracker

Variable	Start	After 1	After 2	After 3	After 4	After 5	Final
Messages Consumed	0	0	100	220	330	440	440
Errors	0	0	0	1	1	1	1
Latency (ms)	0	0	20	25	22	22	22
Metrics Sent	No	Subscription	Batch 1	Batch 2	Batch 3	Periodic	Final

Key Moments - 3 Insights

Why do metrics show zero before subscription?

Before subscribing, the client has not processed any messages, so metrics like messages consumed or errors are zero, as shown in execution_table step 1 and 2.

Why is an alert triggered at step 4?

At step 4, errors count increases to 1, which triggers an alert in the monitoring system, as indicated in the 'Monitoring System Response' column.

How does periodic metrics collection work?

After initial batches, metrics continue to be collected and sent periodically, shown in step 6, ensuring ongoing monitoring without stopping the client.

Visual Quiz - 3 Questions

Look at the execution table, what is the value of 'Errors' after step 3?

A 2

B 1

C 0

D Not collected yet

💡 Hint

Check the 'Metrics Collected' column at step 3 in the execution_table.

At which step does the monitoring system trigger an alert?

A Step 2

B Step 4

C Step 5

D Step 7

💡 Hint

Look at the 'Monitoring System Response' column for alert information.

If the client never subscribes to a topic, what would the 'Messages Consumed' variable show after step 2?

A Exactly 0

B Greater than 0

C Undefined

D Error state

💡 Hint

Refer to variable_tracker for 'Messages Consumed' values before and after subscription.

Concept Snapshot

Client metrics monitoring in Kafka:
- Configure client with metric reporters (e.g., JMXReporter)
- Metrics collected automatically during client operation
- Metrics include messages consumed, errors, latency
- Metrics sent periodically to monitoring system
- Alerts triggered based on metric thresholds
- Continuous monitoring until client stops

Full Transcript

Client metrics monitoring in Kafka involves starting the Kafka client, initializing metrics collection, and subscribing to topics. The client automatically collects metrics such as messages consumed, errors, and latency. These metrics are sent periodically to a monitoring system, which analyzes them and triggers alerts if needed. The process repeats continuously until the client stops. Key moments include understanding why metrics are zero before subscription, recognizing when alerts trigger due to errors, and how periodic collection ensures ongoing monitoring.

Practice

1. What is the main purpose of client metrics monitoring in Kafka?

easy

A. To configure Kafka brokers for better storage

B. To track real-time performance data like message counts and errors

C. To manage user access permissions

D. To backup Kafka data automatically

Solution

1. Step 1: Understand client metrics monitoring

   Client metrics monitoring collects data about Kafka client performance such as message counts, latency, and errors.

2. Step 2: Identify the main purpose

   The main goal is to observe client behavior in real time to improve reliability and troubleshoot issues quickly.

3. Final Answer:

   To track real-time performance data like message counts and errors -> Option B

4. Quick Check:

   Client metrics = performance data tracking [OK]

Hint: Client metrics = real-time client performance data [OK]

Common Mistakes:

• Confusing client metrics with broker configuration
• Thinking metrics manage user permissions
• Assuming metrics handle data backup

2. Which of the following is the correct way to enable client metrics reporting in a Kafka producer configuration?

easy

A. "client.metrics.enabled=yes"

B. "enable.metrics=true"

C. "metrics.recording.level=DEBUG"

D. "metrics.sample.window.ms=30000"

Solution

1. Step 1: Review Kafka producer metric configs

   Kafka uses properties like "metrics.sample.window.ms" to control metrics sampling intervals.

2. Step 2: Identify valid configuration syntax

   "metrics.sample.window.ms=30000" is a valid setting to enable metrics sampling every 30 seconds.

3. Final Answer:

   "metrics.sample.window.ms=30000" -> Option D

4. Quick Check:

   Correct metric config syntax = "metrics.sample.window.ms=30000" [OK]

Hint: Look for valid Kafka metric property names [OK]

Common Mistakes:

• Using non-existent properties like enable.metrics
• Wrong boolean values like 'yes' instead of true/false
• Confusing logging level with metrics config

3. Given the following Kafka consumer metrics output snippet:
{"records-consumed-total": 1500, "fetch-latency-avg": 25, "fetch-errors-total": 0}
What does this tell you about the consumer's performance?

medium

A. The consumer has consumed 1500 records with an average fetch latency of 25 ms and no fetch errors

B. The consumer has 1500 fetch errors and average latency of 0 ms

C. The consumer has not consumed any records yet

D. The consumer is experiencing high fetch errors

Solution

1. Step 1: Interpret each metric value

   "records-consumed-total": 1500 means 1500 messages consumed; "fetch-latency-avg": 25 means average fetch latency is 25 milliseconds; "fetch-errors-total": 0 means no fetch errors.

2. Step 2: Summarize consumer performance

   The consumer is working well, consuming messages with reasonable latency and no errors.

3. Final Answer:

   The consumer has consumed 1500 records with an average fetch latency of 25 ms and no fetch errors -> Option A

4. Quick Check:

   Metrics values = The consumer has consumed 1500 records with an average fetch latency of 25 ms and no fetch errors [OK]

Hint: Check counts and latency values carefully [OK]

Common Mistakes:

• Misreading zero errors as errors present
• Confusing latency with error count
• Assuming no records consumed when count is positive

4. You configured Kafka client metrics but see no metrics reported. Which of the following is the most likely cause?

medium

A. Metrics reporting interval is set too high, delaying output

B. Kafka broker is down

C. Client metrics are disabled by default and need explicit enabling

D. Network firewall blocking Kafka ports

Solution

1. Step 1: Understand default client metrics behavior

   Kafka clients do not always enable metrics reporting by default; explicit configuration is often required.

2. Step 2: Evaluate other options

   Broker down or firewall issues affect connectivity but not metrics reporting directly; reporting interval delays metrics but does not stop them completely.

3. Final Answer:

   Client metrics are disabled by default and need explicit enabling -> Option C

4. Quick Check:

   Metrics disabled by default = Client metrics are disabled by default and need explicit enabling [OK]

Hint: Check if metrics are explicitly enabled in config [OK]

Common Mistakes:

• Assuming broker status affects client metrics directly
• Ignoring need to enable metrics explicitly
• Blaming network issues without checking configs

5. You want to monitor Kafka producer client metrics to detect high message send latency and errors in real time. Which approach is best to achieve this?

hard

A. Configure the producer to expose JMX metrics and use a monitoring tool to collect and alert on latency and error metrics

B. Increase the producer batch size to reduce latency

C. Enable debug logging on the Kafka broker

D. Use Kafka Connect to export metrics to a database

Solution

1. Step 1: Identify how to access client metrics

   Kafka producer clients expose metrics via JMX (Java Management Extensions), which can be collected by monitoring tools.

2. Step 2: Choose the best monitoring approach

   Using JMX with a monitoring tool allows real-time collection and alerting on latency and errors, which fits the requirement.

3. Step 3: Evaluate other options

   Increasing batch size affects performance but not monitoring; debug logging on broker does not monitor client metrics; Kafka Connect exports data but is not the direct way to monitor client metrics.

4. Final Answer:

   Configure the producer to expose JMX metrics and use a monitoring tool to collect and alert on latency and error metrics -> Option A

5. Quick Check:

   JMX metrics + monitoring tool = Configure the producer to expose JMX metrics and use a monitoring tool to collect and alert on latency and error metrics [OK]

Hint: Use JMX metrics with monitoring tools for real-time alerts [OK]

Common Mistakes:

• Confusing performance tuning with monitoring setup
• Relying on broker logs for client metrics
• Using Kafka Connect for direct client metric monitoring