Skip to main content

Min hardware spec

Console requirements

Conduktor is provided as a single Docker container.

Production requirements

For production deployments, ensure: If deploying via Helm, check the install guide for detailed requirements.

Hardware requirements

Use container CGroup limits for resource control.

Minimum

  • 2 CPU cores
  • 3 GB RAM
  • 5 GB disk
  • 4+ CPU cores
  • 4+ GB RAM
  • 10+ GB disk
Configure env variables or get started with Docker.

Memory configuration

Note: RUN_MODE is deprecated. JVM now respects container limits (up to 80%).

Gateway requirements

Conduktor is provided as a Docker image and a Helm chart.

Minimum setup

For light usage per Gateway node:
  • 2 CPU cores
  • 4 GB of RAM
Each Gateway instance should support around 20–30 MB/s of sustained throughput with minimal latency impact.
For medium to high usage per Gateway node:
  • 4 CPU cores
  • 8 GB of RAM
This setup supports ~40–50 MB/s throughput. For production, run at least 3 Gateway instances. Scale horizontally as needed.

Local storage requirements

While Gateway itself doesn’t use local storage directly, some interceptors do. The large message handling interceptor maintains a local disk cache that must be sized for the expected payload volume; see its Local disk cache section for sizing and cleanup behavior.

Scaling Gateway

Gateway is designed to scale horizontally or vertically. It’s mostly CPU-bound unless caching is enabled. Monitor metrics to scale effectively.

Interceptor impact

Interceptors add CPU load. Interceptors doing heavy inspection (e.g. encryption) require:
  • More CPU
  • More memory (recommend 4 GB RAM per CPU)

Kafka requirements

  • Kafka version 2.7.0 or higher
  • Gateway must connect as an admin user with rights to:
    • Manage topics, groups, and cluster info
    • Commit offsets

Deploy in an air-gapped environment

Console and Gateway don’t need internet access to run. License: Console and Gateway verify the license key locally, without contacting Conduktor. AWS Marketplace licensing (CDK_AWS_MARKETPLACE_LICENSE) calls AWS and doesn’t work offline. Images and charts: mirror these images to your private registry: With the Helm charts, set global.imageRegistry (and global.imagePullSecrets if needed) to pull every image from your registry. The chart packages published on helm.conduktor.io include their dependencies, so download them with helm pull and install them from the file or from your internal chart repository. The images used by helm test don’t follow global.imageRegistry. Telemetry: Console, Gateway and the browsers running the Console web UI send product telemetry. Allow it through your egress proxy: the telemetry page lists the destinations. If the traffic is blocked, Console logs Could not upload batch ... Retries exhausted errors, which don’t affect Console. Integrations: some features call external services. For alerts, use webhooks to internal endpoints instead of Slack or Microsoft Teams. For single sign-on (SSO), use your own identity provider instead of Google. The Confluent Cloud and Aiven integrations need outbound access to these providers’ APIs.

Troubleshoot

No. Console and Gateway run without internet access: see deploy in an air-gapped environment.
Yes, the external database is used to store configuration such as users and permissions.