Confluent Certified Administrator for Apache Kafka Free Sample Questions

This is the most effective initial step. By moving the partition leadership to healthy brokers, producers will no longer be dependent on the slow broker for acknowledging writes. This directly addresses the latency issue without changing durability settings, allowing the administrator to investigate the disk I/O problem on the affected broker without impacting production traffic.

The correct format for controller.quorum.voters is a comma-separated string where each entry specifies the node.id, the hostname or IP address, and the controller listener port for each node in the quorum. For example: 1@controller1:9093,2@controller2:9093,....

This property sets the byte-rate threshold for producers. When combined with the user principal, it defines the specific bandwidth limit for that user's produce requests.

This property specifies the user principal to which the quota will be applied. To target 'app-user-1', this must be set to that principal name.

This statement is true. The purpose of unclean leader election is to prioritize availability over consistency. If no in-sync replicas are available to become the new leader, this setting allows an out-of-sync replica to be elected. This new leader may not have all the messages the old leader had, leading to data loss.

Log compaction works by retaining the latest message for each unique key. If messages are produced with null keys, Kafka cannot determine which records to compact. Messages with null keys are not eligible for compaction and will be retained until their segment is deleted by time or size-based retention, if also configured. This is the most common reason for compaction appearing not to work.

When a worker in a distributed Kafka Connect cluster fails or leaves the group, the group coordinator triggers a rebalance. During this rebalance, all tasks from the failed worker (and potentially other workers) are redistributed among the surviving members of the cluster, ensuring that data processing continues with minimal interruption.

This is the correct command. It specifies the operation (Read), the principal (user:data-science), the resource name (ds-), and crucially, sets the --resource-pattern-type to prefixed. This tells the authorizer to match any topic whose name begins with ds-. Don't forget that you also need to grant Read access to the consumer group.

The max.poll.interval.ms setting defines the maximum time allowed between calls to the poll() method. If the consumer takes longer than this to process the records from a single poll (6 minutes > 5 minutes), the broker's group coordinator will assume the consumer has failed and kick it out of the group, triggering a rebalance. The logs might not show it immediately, but this constant rebalancing and reprocessing is the direct cause of the lag.

The kafka-topics.sh --describe command provides detailed information about topics, including the leader, replicas, and the In-Sync Replica (ISR) set for each partition. By comparing the list of replicas with the ISR list, an administrator can identify any partitions that are under-replicated.

The broker.rack property is used to enable Kafka's rack awareness feature. By setting this property on each broker to the name of the availability zone it resides in (e.g., us-east-1a), Kafka will make a best effort to spread replicas of a partition across different racks (AZs), preventing a single AZ failure from causing data loss or unavailability.