CMAK vs Confluent Control Center
ComparisonsCMAK vs Confluent Control Center, compared
Key takeaway
CMAK requires a ZooKeeper ensemble, which Kafka 4.0 has none of, so a KRaft cutover ends it the day it completes. Control Center requires the proprietary Confluent Metrics Reporter in the broker classpath, so it cannot watch Amazon MSK, Redpanda or Aiven. Neither has a price you can compare, and they are not the same kind of product: CMAK administers and has never browsed messages, and Control Center’s Streams and ksqlDB views have no open-source equivalent. Kpow by Factor House is licensed per cluster at a published price.
What is CMAK?
CMAK, the Cluster Manager for Apache Kafka and originally Kafka Manager, is free software built at Yahoo and released under Apache 2.0. It is written in Scala on the Play framework, and it requires a direct connection to a ZooKeeper ensemble to function at all. Its scope is Kafka cluster management rather than the data plane, which makes it a Kafka management console rather than a viewer.
- registering and monitoring clusters from one multi-cluster view
- partition reassignment and preferred-replica election
- dynamic topic configuration, partition creation and replica change
- optional JMX polling at broker and topic level
The last stable release, 3.0.0.6, was tagged on 29 April 2022. The most recent commit on master landed on 12 December 2022, and the repository has seen no push since 2 August 2023. Some 522 issues stand open with nobody triaging them. The only Kubernetes deployment path is a third-party Helm chart, archived and read-only since June 2023.
What is Confluent Control Center?
Control Center is a web management and monitoring interface bundled with Confluent Platform, Confluent’s commercial Kafka distribution. It is closed-source, licensed as part of the platform, and not available standalone. It gives one dashboard over brokers, topics, consumer groups, Kafka Connect workers, Schema Registry, ksqlDB and Kafka Streams topologies.
The prerequisite decides most evaluations: it requires the proprietary Confluent Metrics Reporter JAR in the broker classpath, and that JAR cannot be installed on Amazon MSK, on Redpanda or on Aiven. Kafka Streams topology visualisation and native ksqlDB development have no equivalent in any open-source Kafka UI, which is the capability the licence genuinely buys.
- Legacy architecture: a Kafka Streams metrics pipeline, shipped with Confluent Platform 7.x and earlier.
- Next generation: Prometheus-based, generally available with Confluent Platform 8.0 in May 2025.
- What changed: startup fell from 15 to 50 minutes to roughly one, supported scale rose from 120,000 to 400,000 replicas, and the separate metrics cluster went away.
What is the official 2026 pricing of CMAK and Confluent Control Center?
Neither of these has a price ladder to read. CMAK has no price at all, so what it costs is carry. It is distributed as source needing an sbt and Scala build, with build friction on current toolchains including an OpenJDK 17 compatibility problem. No container image is published officially. Long-running instances hang after 20 to 30 days with a RejectedExecutionException from thread-pool exhaustion, reported in 2018 and never addressed, and larger clusters with JMX polling need manual thread-pool tuning by formula.
Control Center is not sold separately, so it has no price of its own either. It is bundled with Confluent Platform under an enterprise licence that is not published, and Control Center, multi-tenancy support and encryption each carry cost above the base licence. The Platinum support tier is not available for this product. Next-generation sizing is 4 cores, 8 GB of RAM and 200 GB of storage, preferably SSD, for clusters up to 100,000 replicas, rising to 8 cores and 16 GB above that, with storage assuming 15 days of metrics retention. For fifty engineers the licence is a line item defended every year against the Kafka monitoring stack the team probably already runs in Prometheus and Grafana.
Where does each one run out?
CMAK’s ZooKeeper dependency is the whole clock. Kafka 4.0 operates entirely without ZooKeeper, so the morning a KRaft cutover completes, the admin console stops working, for the team that is mid-migration and least able to lose its instruments. The maintainer acknowledged this in 2022 and nothing has shipped since. MSK, Confluent Cloud, Aiven and Redpanda Cloud either lock down the ZooKeeper endpoint or no longer expose one.
- Data plane: no message browsing, no Schema Registry integration and no Kafka Connect management.
- Freshness: reads come from an internal cache, so hours of catch-up replication can show as complete in seconds.
- Access control: LDAP basic auth and coarse global feature flags, with no per-user, per-cluster or per-topic granularity, and no audit log.
- Transport: the LDAP integration is unencrypted unless SSL is configured by hand, which the README states.
Control Center is scoped to one distribution, and inside that scope it is heavier than it looks. It cannot monitor Amazon MSK, Redpanda or Aiven at all, and MSK’s native IAM authentication is not supported.
- Single sign-on: OIDC only on self-managed deployments. SAML is Confluent Cloud only, so a SAML-only identity provider has no supported path.
- RBAC: access rules cap at 10,000 per cluster, and metrics cannot be sent without enabling full management. See RBAC for Kafka.
- GitOps: changes apply directly to cluster state rather than through a Git-managed manifest.
- Legacy interceptors: roughly 50 internal topics added to broker metadata, and at high consumer group counts they can overwhelm the product.
Kafka Streams can also enter a rebalancing loop on startup and leave the interface on a loading spinner for 20 to 30 minutes. Moving from legacy to next generation is a migration rather than an upgrade: historical metrics do not carry across, and the guidance is 7 to 15 days running both in parallel.
Which should you pick?
Take Control Center if:
- you are already on Confluent Platform
- Kafka Streams topology visualisation is a requirement
- ksqlDB development integration is a requirement
Keep CMAK, with a date on it, if:
- the cluster is still on ZooKeeper
- the daily work is partition reassignment and preferred-replica election
- nobody on the team needs to read a message from the interface
If you are not on Confluent Platform, this is a distribution purchase rather than a console purchase, and it should be evaluated as one. These two sit on the same lists of the best Kafka management tools, compared feature by feature, and the feature comparison is not what decides it: a team on Amazon MSK, Redpanda or Aiven can run neither, and the reassignment work CMAK does well is not covered by the viewers on any list of the best free Kafka UI tools.
Kpow: not gated by what the cluster is
Both of these decide what they can see before they see a single broker: CMAK by whether a ZooKeeper ensemble is reachable, and Control Center by whether the cluster is running Confluent Platform with the proprietary Metrics Reporter JAR already in the broker classpath. Neither answer travels — a KRaft cutover ends CMAK on the day it completes, and a cluster on Amazon MSK, Redpanda or Aiven was never eligible for Control Center at all. Kpow by Factor House connects to self-managed Kafka, MSK, Confluent Cloud, Redpanda, Aiven and Instaclustr from the same install, runs as a single stateless container with no external database, and stores its own telemetry in internal Kafka topics rather than a side system. It is licensed per cluster at a published price, so which cluster you run stops being the first question.
The cluster you actually have doesn’t have to disqualify the console you use to run it. See what Kpow reaches and settle the gating question for good.
Sources
- KIP-833: Mark KRaft as Production Ready
- KIP-500: Replace ZooKeeper with a Self-Managed Metadata Quorum