Monitor Kafka Consumer Lag in Confluent Cloud

Consumer lag is the difference between the rate at which producers write records to an Apache Kafka® topic and the rate at which consumers read them. When consumers fall behind, unprocessed records accumulate. This backlog is called consumer lag. Consumer lag has two components:

Offset lag: The gap between the latest produced offset and the last committed consumer offset.
Consumer latency: The time delay between production and consumption.

In Confluent Cloud, you can monitor consumer lag using the Metrics API, the Confluent Cloud Console, the Kafka Admin API, the Confluent CLI, or the REST API.

To identify the specific cause of lag, you can use different tools depending on the level of detail you need. Use the Cloud Console and Metrics API for overall lag metrics, and use Java client metrics and the Kafka Admin API to monitor offset lag specifically. For more information, see Use the Metrics API to monitor Kafka consumer groups, Use the Confluent Cloud Console to monitor Kafka consumer lag, Use Java client metrics to monitor Kafka offset lag, and Use Kafka Admin API to monitor offset lag.

Choose a monitoring method
Method	Best for	Granularity	Setup
Metrics API	Production monitoring, alerting, dashboards	Topic + consumer group (partition with query)	API call or integration
Cloud Console	Quick visual checks, ad-hoc debugging	Consumer group + partition	None (built-in)
Java client metrics	Application-level monitoring	Per-consumer instance	JMX or metrics reporter
Kafka Admin API + CLI	Programmatic offset tracking, scripting	Partition-level	CLI + API key
Confluent CLI	Quick consumer group lag checks (Dedicated clusters only)	Consumer group + partition	CLI + API key
REST API	Programmatic consumer group access (Dedicated clusters only)	Consumer group	API key

Use the Metrics API to monitor Kafka consumer groups

Use the Metrics API to monitor how consumer lag changes over time. To monitor at the topic and consumer group level of detail, you can use a supported integration. To view data at the more detailed consumer and partition level, you can begin from the example query.

This metric differs from the kafka-consumer-groups command output in two ways:

Consumer groups do not update lag values during rebalancing.
Consumer groups with no active consumers return no result.

For an example of the rebalance behavior, consider a consumer group that has two active consumers that are lagging behind by three and four records each. Metrics API returns respective values of three and four for those consumers. A rebalance begins and producers send more records to the topic but those records cannot be consumed until the rebalance completes. The rebalance takes two minutes. For those two minutes, the value for lag does not change (three and four) despite 20 additional records being sent to the topic during those two minutes. When rebalance completes, Metrics API returns respective values of 13 and 14 for those consumers. Metrics API does not update consumer lag values during rebalance.

Now consider an example where two consumers are unresponsive and no longer send heartbeats. Metrics API no longer returns data for this consumer group. Metrics API does not expose consumer lag for topics and partitions that do not have an active consumer.

Important

Metrics API does not include consumers that use the assign() method. This is because the coordinator of a consumer group does not manage consumer assignment for consumers assigned to topics and partitions using assign().

Monitor consumer group rebalancing

Along with monitoring consumer lag, also monitor consumer group rebalance events. Frequent rebalance events or long individual rebalance events could result in lower productivity of your applications and could result in service disruption. This is more likely when using consumer groups with the classic group protocol, which is the only option available before Kafka version 4.0.

To monitor rebalance events, use the io.confluent.kafka.server/max_pending_rebalance_time_milliseconds metric, which shows the maximum pending rebalance time in milliseconds among all members of a consumer group. The value shows how long a consumer group is in the rebalancing state. A value of zero for this metric indicates that the consumer group is in a state other than rebalancing. Here is an example response when querying this metric:

 {
     "data": [
      {
          "metric.consumer_group_id": "test-group",
          "metric.group_protocol": "CLASSIC",
          "timestamp": "2025-11-14T16:00:00Z",
          "value": 0.0
      },
      {
          "metric.consumer_group_id": "test-group",
          "metric.group_protocol": "CLASSIC",
          "timestamp": "2025-11-14T16:01:00Z",
          "value": 14594.0
      },
      {
          "metric.consumer_group_id": "test-group",
          "metric.group_protocol": "CLASSIC",
          "timestamp": "2025-11-14T16:02:00Z",
          "value": 0.0
      }
    ]
}

In this example, the consumer group had a rebalance event at 2025-11-14T16:01:00Z which lasted for about 14.5 seconds. Previously, this specific consumer group was empty and the rebalance event occurred when a consumer joined the group. After the rebalance event, the consumer group went to being stable. This metric does not discern whether the group was empty or stable during periods where the value shows 0.0. However, you can gather that from situational context or more explicitly through client logs. Depending on the situation, you can use this metric to understand the following:

How long individual rebalance events last.
How many rebalance events occurred in a given time frame.
When a certain group was last seen not in a rebalancing state.

To help reduce the impact of rebalance events in general, use the new consumer group protocol type, which was introduced in Kafka version 4.0. For more information on the new group protocol and how it can help reduce the impact of rebalancing, see groups and rebalance protocols.

Use the Confluent Cloud Console to monitor Kafka consumer lag

Cloud Console uses the Metrics API to monitor consumer lag and provides output that is different from kafka-consumer-groups. For more information, see Use the Metrics API to monitor Kafka consumer groups.

You might notice clients that use a naming convention similar to this: consumer-nnnn. These clients represent the Confluent Cloud interface itself.

To monitor consumer lag with Cloud Console:

Select your cluster name.
Choose Clients and select Consumer Lag. A list of consumer groups displays.
- The consumer group list displays information about your consumer groups, including the group ID, status, and protocol. You can also find values for group lag, and number of consumers, topics, and partitions. For more information, see Kafka Consumer for Confluent Cloud.
- Cloud Console limits the number of consumer groups shown in Consumer lag if there are many.
- You can export consumer lag data as a CSV file. For clusters with large amounts of data, you might see a placeholder value like -999 in the currentOffset field. This occurs because Cloud Console cannot display all the data immediately, and the offsets are not yet available. For large payloads, consider exporting consumer lag data using the Confluent CLI or Confluent Cloud APIs.
Select a consumer group from the list to see lag details for that group.

For more information on creating a consumer, see Quick Start for Confluent Cloud.

Use Java client metrics to monitor Kafka offset lag

Offset lag is the difference between the latest offset available in a Kafka topic partition and the offset that a consumer group has consumed. This value indicates how far behind the consumer group is from the latest available data.

You can monitor the records-lag-max metric from the Java consumer.

Use Kafka Admin API to monitor offset lag

You can monitor offsets by using the Kafka Admin API and the associated command-line tool, which enables accessing lag information programmatically. For more information, see AdminClient Configurations.

Prerequisites

Access to Confluent Cloud.
Java version 1.7.0_111 or later, 1.8.0_102 or later, and 1.9.
A web browser.
Confluent Platform is installed.

Create a client properties file to hold the Confluent Cloud configuration. In this example, it is named client_ssl.properties. This file should contain the Confluent Cloud client configurations. You can find this information in the CLI and client configuration tab of Confluent Cloud Console. Configure this example for your environment:
```
ssl.endpoint.identification.algorithm=https
sasl.mechanism=PLAIN
sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required \
username="<cluster-api-key>" \
password="<cluster-api-secret>";
security.protocol=SASL_SSL
```
Note
Use a Confluent Cloud API key (resource-scoped for the Kafka cluster) and secret for the username and password configs. Using a Confluent Cloud API key with the wrong resource scope causes authentication to fail.
Set the BOOTSTRAP_SERVERS variable to the Confluent Cloud cluster bootstrap URL. You can find this value by clicking Cluster settings in the Cloud Console.
```
BOOTSTRAP_SERVERS="<bootstrap-url>"
```
From the Confluent Platform installation home, list the consumer groups. Pass Confluent Cloud properties with the --command-config argument. You must provide a bootstrap server to the script.
```
./bin/kafka-consumer-groups --bootstrap-server ${BOOTSTRAP_SERVERS} --command-config \
client-ssl.properties --list  _confluent-healthcheck  example-group
```
If the previous command causes a timeout error, try increasing the timeout to 10 seconds by using the --timeout option, for example, --timeout 10000.

For each consumer group, check its offsets using this command. This command only shows information about consumers that use the Java consumer API (that is, non-ZooKeeper-based consumers).

./bin/kafka-consumer-groups --bootstrap-server ${BOOTSTRAP_SERVERS} \
 --command-config /tmp/client.properties --describe --group _confluent-healthcheck

Your output should resemble:

TOPIC                          PARTITION  CURRENT-OFFSET  LOG-END-OFFSET  LAG        CONSUMER-ID                                       HOST                           CLIENT-ID
_confluent-healthcheck         0          13164704        13164773        69         healthcheck-agent-bf8d1655-63a6-4061-b680-0f11cdf182e5/100.96.67.0                   healthcheck-agent
_confluent-healthcheck         1          13161581        13161650        69         healthcheck-agent-bf8d1655-63a6-4061-b680-0f11cdf182e5/100.96.67.0                   healthcheck-agent
_confluent-healthcheck         2          12229509        12229578        69         healthcheck-agent-bf8d1655-63a6-4061-b680-0f11cdf182e5/100.96.67.0                   healthcheck-agent
_confluent-healthcheck         3          86              86              0          healthcheck-agent-bf8d1655-63a6-4061-b680-0f11cdf182e5/100.96.67.0                   healthcheck-agent
...

The fourth column shows the lag, the difference between the last committed offset and the latest offset in the log.

Known kafka-consumer-groups issues

You might encounter the following error, or the command might fail with a TimeoutException:

Error: Executing consumer group command failed due to Failed to construct kafka consumer

In either case, raise the request.timeout.ms configuration in client_ssl.properties (which defaults to 5000 in the kafka-consumer-groups command line tool). For example:

request.timeout.ms=60000

Best practices for monitoring consumer lag

Set alerting thresholds on consumer lag. Alert when lag exceeds a defined number of records or grows continuously for a sustained period. Use the Metrics API or a third-party integration to configure alerts.
Monitor lag trends, not absolute values alone. A small, steady lag is normal. Growing lag over time indicates a consumer that cannot keep up with the production rate.
Check offset lag and consumer latency. Offset lag shows how many records are unprocessed. Consumer latency shows the time delay. Together, they help you distinguish between slow consumers and stuck consumers.
Monitor rebalance frequency and duration alongside lag. Frequent or long rebalance events cause lag spikes. Use the max_pending_rebalance_time_milliseconds metric to detect them.
Use the consumer group protocol to reduce rebalance impact. The consumer group protocol, introduced in Kafka version 4.0, reduces the duration and frequency of rebalance events compared to the classic protocol.