Migration guide

Moving from Confluent to managed Apache Kafka

Move Confluent Cloud or Confluent Platform to Apache Kafka run in our cloud, your cloud account or your data center, with MirrorMaker 2 keeping both clusters live.

  • Operated by AceMQ
  • 160+ customers
The migration in phases

A Confluent to Kafka migration uses Apache Kafka's MirrorMaker 2, which replicates topics with their configuration, consumer groups with offsets and ACLs between clusters, so a Confluent cluster and a new Apache Kafka cluster run side by side. Schema Registry data and connectors are migrated separately.

Which migration method fits

MirrorMaker 2 is the Apache Kafka project's tool for copying data between clusters, and it keeps both live. The Schema Registry and the connectors are separate parts of the estate and have their own steps.

Illustrative: each phase is planned and rehearsed with your team before traffic moves.
Part of the estateDocumented methodBoth sides liveSource
Topics, offsets and ACLsMirrorMaker 2, built on Kafka Connect, replicates topics (data and configuration), consumer groups including offsets, and ACLs, and preserves partitioning. It detects new topics and partitions automatically.YesApache Kafka docs
ConnectorsConnectors run on Kafka Connect workers. Those managed by Confluent Cloud are recreated as connector configurations on self-run workers.Per connectorApache Kafka docs
SchemasConfluent documents subject-level migration through the Schema Registry API, which preserves schema ID and version information, with the destination in import mode.YesConfluent docs

Sources: Apache Kafka and Confluent documentation, linked in each row; checked 30 Oct 2026.

How the methods were chosen: MirrorMaker 2 is the cross-cluster tool in Apache Kafka itself, so it works with any target that speaks the Kafka protocol. Confluent's own replication products are not used, because the target is open source Apache Kafka.

Check these before you start

MirrorMaker 2 changes topic names by default, offsets move through a documented sync, and schema IDs must survive the move. These three decide how the applications are switched.

Topic names.
By default, replicated topics take the source cluster name as a prefix, and a replication policy can change this (Apache Kafka docs). The applications either read the new names or the policy keeps the old ones, and the design chooses.
Consumer offsets.
MirrorMaker 2 replicates consumer groups including offsets so applications can migrate between clusters (Apache Kafka docs). Automated offset sync is described in KIP-545.
Exactly once.
Exactly-once semantics are supported for dedicated MirrorMaker clusters as of version 3.5.0 (Apache Kafka docs).
Schema IDs.
Messages written with a schema registry carry the schema ID, so the IDs must be the same on the new registry. The Confluent docs describe migration that preserves schema ID and version.
Network path.
The MirrorMaker 2 workers must reach the Confluent bootstrap endpoint, over its public endpoint or a private link, and the credentials are mapped to the new cluster's authentication. Both are design items.

Sources: Apache Kafka and Confluent documentation, linked above; checked 30 Oct 2026.

The migration in phases

The new cluster is built and mirrored while Confluent keeps serving. Consumers and producers move only after the mirror has caught up, and the old cluster stays until the new one is verified.

  1. Review

    Topics, partitions, retention, consumer groups, ACLs, connectors, schemas, throughput, and how clients authenticate.

  2. Build the new cluster

    Brokers, storage, Connect workers and the schema registry come from the design, in the place you choose.

  3. Mirror

    MirrorMaker 2 starts a replication flow from the Confluent cluster to the new one. Lag is watched until it is low.

  4. Move schemas and connectors

    Schemas migrate with IDs preserved, and connectors are recreated on Connect workers and tested.

  5. Move consumers, then producers

    Consumers can read the mirrored topics first while producers still write to Confluent. Producers follow once the lag reaches zero.

The timeline and runbook for Confluent to Apache Kafka migrations are written in the design step, from your data volume, traffic and change windows, and you approve them before the cutover is scheduled.

Cutover and rollback

Order matters more than speed. Consumers move first on mirrored data, producers second, and Confluent stays in place and mirrored until the new cluster has run normally.

  1. Mirror lag is near zero and consumer group offsets are in step.
  2. Consumers connect to the new cluster and read the mirrored topics.
  3. Producers switch to the new cluster, and the last records on Confluent are mirrored.
  4. Confluent is retired only after the new cluster has served normal traffic.

Source: the Apache Kafka geo-replication documentation (Apache Kafka docs); checked 30 Oct 2026.

How OSStend runs the move

We run the migration and own the cutover, with both sides live until you switch. Zero downtime is the plan.

Managed Kafka from OSStend includes Kafka Connect and a schema registry, operated by AceMQ engineers in our cloud, your cloud account or your data center. The reasons teams leave Confluent are set out on the Confluent Cloud vs managed Kafka page, and the migration hub lists the other moves.

Operated by AceMQCompany figures, as published on acemq.com
160+customers served by AceMQ, which operates OSStend
10+years running open source data and messaging
24/7on-call engineers, not a ticket queue

AceMQ names a customer only with that customer's written permission, so most of this record stays anonymous. Ask for a similar workload in the 45-minute review.

When this move fits, and when it does not

The move fits when cost, location or control is the reason to leave Confluent and the estate uses Apache Kafka features. It does not when the applications depend on Confluent-only products.

Contact us if

  • Cost at scale, data location or control over the cluster is the reason to leave.
  • The estate uses Apache Kafka, Kafka Connect and a schema registry.
  • Producers and consumers must keep running while the move happens.
Talk to an engineer

Not a fit if

  • The applications depend on Confluent-only products with no open source replacement.
  • You want the full Confluent platform and accept the cost.
  • You want to run the move yourselves: MirrorMaker 2 works without OSStend.

Questions about moving off Confluent

Can Confluent Cloud be migrated to Apache Kafka without downtime?

MirrorMaker 2 replicates topics continuously, so Confluent and the new cluster stay live while consumers and producers move in stages (Apache Kafka docs, checked 30 Oct 2026). Connectors and schemas are migrated alongside.

Does MirrorMaker 2 copy consumer offsets?

Yes. It replicates consumer groups including offsets to migrate applications between clusters, and offset sync can be automated (Apache Kafka docs, checked 30 Oct 2026).

What happens to schemas and connectors?

They are separate from MirrorMaker 2. Schemas move through the Schema Registry API with IDs preserved (Confluent docs, checked 30 Oct 2026), and connectors are recreated on Kafka Connect workers.

Talk to an engineer

Plan your Kafka move with an engineer

45 minutes with an engineer: the current setup is reviewed against the documented method and the cutover is sketched. No ticket queue.

  • A design sketch for your workload, not a plan menu
  • A zero-downtime cutover path for what you run today
  • A price for that design, after the review
See pricing →

Your request goes to the OSStend team. No newsletter sign-up.

Get a quoteTalk to an engineer
2Engines
3Where should it run?
4What do you need?

Your request goes to the OSStend team. No newsletter sign-up.