Kafka is known as the best streaming platform, it has the capability to handle data in a pull mechanism, making it highly performant. Kafka provides capabilities to segregate data via topics, it supports binary data and text data. It has the capability to distribute data to partitions, a notion of shards. Kafka has consumers - Kafka streams which are built on the principles of Kafka to give better performance. We considered replacing Storm with Kafka streams and could achieve all the required functionalities. Kafka provides various tune-ups to achieve HA. Review collected by and hosted on G2.com.

One thing I observed is we cannot switch off replication in a multi-node setup. Suppose I want only load distribution, but it's okay if I lose data. I don't want any replication. If any partition-assigned node goes down, Kafka should identify and reassign the partitions at some point to other nodes; instead, it goes on toss expecting ISR to be available. Another point is when any node goes down, it takes a while to reassign the primary partitions and make Kafka available, though there is a config to set the time to trigger rebalancing, usually it kicks the rebalancing, but the rebalancing itself takes a lot of time. Review collected by and hosted on G2.com.

The nature of message streaming is guaranteed to all consumers who subscribed and good thing is the message will never missed which guarantees data delivery Review collected by and hosted on G2.com.

No issues with Kafka in terms of topics and streaming and couldn’t any drawbacks Review collected by and hosted on G2.com.

The way we can scale it, as opposed to rabbitmq, the persistent intermediate storage with ksql, and paralleling data by topics Review collected by and hosted on G2.com.

The official sink and source connectors, from debezium have issues with some advanced data types, especially arrays and son, when reading from postgres WAL files. Review collected by and hosted on G2.com.

Apache Kafka is a very fast and scalable messaging system. I used it for all the microservices asynchronous communication channels. It is extremely easy to setup and configure according to your needs. It also has sdks to help using it with wide range of technology stack. It also has a central broker which governs the master instance and takes care of failure and scaling. Review collected by and hosted on G2.com.

The fact that it broker is not inbuilt and requires zookeper. But its not a dealbreaker. Also that you do not have a dedicated gui for tracking and maintenance purposes. Review collected by and hosted on G2.com.

I like the simplicity of setup of Kafka broker in cluster mode. It permits to have a three powerful Kafka brokers by using even a single virtual low cost machine. The interaction with the cluster is quite simple and you can customize the topics functionalities by setting replication factors and main copy broker. I like also the flexibility of Kafka. It can be used for managing high quantity of messages from multiple sensors, but also to stream a single big video to multiple subscribers. Review collected by and hosted on G2.com.

I have not found anything that I dislike until now. Maybe the confluent support can be improved, since they offer only support for payed Confluent licenses and not for on-premises Kafka installation. Review collected by and hosted on G2.com.

Very fast messaging platform - consumer groups are a great feature - So is replication. Kafka streams are fantastic. Review collected by and hosted on G2.com.

Whatever I have used worked well. Some more standard monitoring tools should be part of the basic setup. Review collected by and hosted on G2.com.

Ability to use Kafka as a database, capability to handle large volumes of data efficiently. Review collected by and hosted on G2.com.

Very few monitoring tools, not all scripts available for monitoring. Review collected by and hosted on G2.com.

Consumer group is one of my favourite feature which let multiple consumers to consume at the same time but avoid reading duplicates. Review collected by and hosted on G2.com.

It’s highly configurable, I mean with configuration settings we can manage Kafka. This is both good and bad. It’s really easy to set up, but if we miss something, we get screwed. One good example: Our Kafka consumer will not poll until it processes all the currently running jobs. In this case, the running jobs take up to an hour. So, Kafka consumer will not do any activities for an hour. In this case, zookeeper assumes my Kafka consumer died and recreates the new one and reads the same message which processed already. This was finally a config setting. Review collected by and hosted on G2.com.

We were looking for an open-source solution to perform stream processing. When our team found this, it was immediately accepted to be used as our permanent solution to achieve the goal we had. Review collected by and hosted on G2.com.

We haven't found any issues so far in Kafka. Review collected by and hosted on G2.com.

Low Latency, High Throughput and replication Review collected by and hosted on G2.com.

Apache Kafka does not support wildcard topic selection. Instead, it matches only the exact topic name. It is because selecting wildcard topics make it incapable to address certain use cases. Review collected by and hosted on G2.com.