Build Event Streams (managed Kafka) configurations with topics, schema registry, and consumer groups.
Last verified: May 2026
Build Event Streams (managed Kafka) configurations with topics, schema registry, consumer groups, networking, and mirroring.
Required Fields
instanceNameplantopicsOutput will appear here...The builder collects instance name, region, plan, topic definitions (name, partitions, retention, cleanup policy), and schema registry subjects. It validates the plan-throughput combination against current Event Streams limits and emits `ibm_event_streams_instance` and `ibm_event_streams_topic` Terraform resources. Schema registry subjects are emitted as inline configuration on the instance.
IBM Event Streams is managed Apache Kafka on IBM Cloud, with topics, partitions, consumer groups, and a built-in schema registry. The Event Streams Config Builder generates a complete instance configuration including plan selection (Lite, Standard, Enterprise), topic definitions with partitions and retention, and schema registry subjects. Output is Terraform-ready and matches the parameters expected by IBM Cloud Kafka resources.
Your application has been on self-managed Kafka in EC2 and the on-call rotation is tired of broker maintenance. You spec out an Event Streams Enterprise instance with topics matching the existing workload, generate the Terraform, and apply. After a controlled cutover (dual-write for a week, then flip consumers), the self-managed brokers are decommissioned. On-call burden drops by 60% — Event Streams handles broker patching, rebalancing, and capacity.
Always use the schema registry for production topics. Untyped Kafka topics become unmanageable as producers and consumers evolve — schema registry enforces compatibility and gives you a documented contract.
Don't size partitions for current load; size for expected peak in the next 12 months. Re-partitioning is operationally painful (consumers must be paused or reset) — better to over-partition slightly than to need to scale up under fire.
Partition count sets the upper limit on consumer parallelism — a consumer group can have at most as many active consumers as the topic has partitions. Pick partition count based on your peak throughput target and target latency. More partitions = more parallelism, but also more overhead. Common starting points: 6 for low-throughput topics, 24 for high-throughput, scale up only when you measure a bottleneck.
Longer retention means more storage cost and a longer history for consumers to replay. Shorter retention means lower cost but consumers that fall behind may lose data permanently. Match retention to your worst-case consumer lag: a daily-batch consumer needs at least 30 hours of retention; a real-time consumer can do with 12. Default 7 days is conservative and often wrong for cost.
Was this tool helpful?
Disclaimer: This tool runs entirely in your browser. No data is sent to our servers. Always verify outputs before using them in production. AWS, Azure, and GCP are trademarks of their respective owners.