Free Confluent Certified Administrator for Apache Kafka CCAAK Exam Questions

10 partitions 25 GB/day = 250 GB total per day for the topic (primary data).

With a replication factor of 3, there are 3 full copies of the data: 250 GB 3 = 750 GB total across the entire cluster.

The cluster has 3 brokers, and Kafka tries to distribute replicas evenly among them: 750 GB 3 brokers = 250 GB per broker on average.

However, due to replication, some partitions have leaders and followers, so there's some overlap and not-perfect distribution. Each broker stores approximately 2/3 of the total topic data (since each broker holds replicas for around 2/3 of the partitions).

2/3 750 GB = 500 GB, but this is shared, so each broker ends up storing ~300 GB of replicated data, including its share of leaders and followers.

When a broker goes down, the Controller detects the failure and triggers a leader election for all partitions that had their leader on the failed broker. The leader is chosen from the in-sync replicas (ISRs) of each partition.

A Sink Connector reads data from a Kafka topic and writes it to an external data system, such as a database, file system, or cloud service.

When using Confluent Replicator (or MirrorMaker), offsets are not preserved between the source and destination Kafka clusters. Messages are replicated based on content, but they are assigned new offsets in the DR (destination) cluster. Therefore, offset 1002 on the source and offset 1002 on the DR cluster likely refer to different messages, which is expected behavior.

The connector is currently configured with 'tasks.max': '1', which means only one task is handling all tables (users, payment, orders). This can create a bottleneck and lead to lag. Increasing tasks.max allows Kafka Connect to parallelize work across multiple tasks, which can pull data from different tables concurrently and reduce lag.

A Leader of -1 indicates that no broker is currently the leader for that partition. This usually happens when the only replica for that partition is unavailable, often due to the associated broker (in this case, Broker 1) failing or being offline. Kafka cannot elect a leader if no replica is in the in-sync replica (ISR) list, which leads to leader = -1.

In a Kafka cluster, only one broker acts as the Controller at any given time. The Controller is responsible for managing cluster metadata, such as partition leadership and broker status. Even if the cluster has multiple brokers (in this case, four), only one is elected as the Controller, and others serve as regular brokers. If the current Controller fails, another broker is automatically elected to take its place.

Kafka checks permissions in the following sequence:

1. Super Users: If the user is a super user (defined via super.users), access is granted immediately.

2. Allow ACL: If there is a matching Allow ACL, Kafka proceeds to the next step.

3. Deny ACL: If there is a matching Deny ACL, access is denied (even if an Allow exists).

4. Deny: If no matching ACLs are found, access is denied by default.

This order ensures that super users bypass ACLs, denials override allows, and default is deny.

Kafka topics are split into one or more partitions to enable parallelism and scalability.

Each partition belongs to exactly one topic; it cannot span multiple topics.

connect-configs stores connector configurations.

connect-status tracks the status of connectors and tasks (e.g., RUNNING, FAILED).

connect-offsets stores source connector offsets for reading from external systems.