# Strimzi 0.45.1

2025-06-27 · [Original release notes](https://github.com/strimzi/strimzi-kafka-operator/releases/tag/0.45.1) · [Full summary](https://ratatosk.io/en/releases/strimzi/0.45.1)

Security 2 · Breaking changes 3 · Important fixes 3 · Enhancements 3 · Dependency updates 8 · Other changes 10

## Check before upgrading

### Prerequisites

- **Strimzi 0.45 drops support for Kubernetes 1.23 and 1.24**: Kubernetes 1.25 is the minimum supported version.
  > ⚠️ **Important: Strimzi 0.45 supports only Kubernetes 1.25 and newer. Kubernetes 1.23 and 1.24 are not supported anymore.**
- **Last Strimzi release supporting ZooKeeper clusters and MirrorMaker 1**: Upgrading to Strimzi 0.46 or newer requires migration to KRaft and MirrorMaker 2.
  > * **Strimzi 0.45 is the last minor Strimzi version with support for ZooKeeper-based Apache Kafka clusters and MirrorMaker 1 deployments.**
  >   **Please make sure to [migrate to KRaft](https://strimzi.io/docs/operators/latest/full/deploying.html#assembly-kraft-mode-str) and MirrorMaker 2 before upgrading to Strimzi 0.46 or newer.**

  - KRaft deployments require `Kafka` and `KafkaNodePool` custom resources
  - Kafka version compatibility with KRaft metadata
  - Static controller quorums used for all deployments: Apache Kafka does not support migration between static and dynamic controller quorums, so existing KRaft clusters using static quorums must continue using them.
  - Controller quorum scaling cannot add or remove controller node pools
  - Dynamic controller quorum scaling cannot add controller roles to existing node pools
  - Controller quorum scaling cannot remove controller roles from existing node pools
  - Dynamic controller quorum scaling cannot scale controller node pools
  - Controller node pool renames are unsupported with dynamic quorum scaling
  - Static controller quorums block storage type changes for controller-role node pools: With non-JBOD storage, replacing an old pool with a new pool using the desired storage type also requires scaling operations that are unsupported.
  - Broker roles on controller nodes as a scaling workaround: This avoids scaling controllers, but requires reassigning more data and may temporarily affect cluster performance.
  - KRaft migration required before upgrading ZooKeeper-based clusters
- **Direct upgrades from Strimzi 0.22 or earlier no longer supported**: The supported path is through an earlier Strimzi version, with the CRD resources converted.
  > **Direct upgrade from Strimzi 0.22 or earlier is not supported anymore!** You have to upgrade first to one of the previous versions of Strimzi. You will also need to convert the CRD resources. For more details, see the [documentation](https://strimzi.io/docs/operators/0.45.1/deploying.html#assembly-upgrade-str).

### Upgrade guide

- **Upgrade instructions documentation**
  > See the [documentation for upgrade instructions](https://strimzi.io/docs/operators/0.45.1/deploying.html#assembly-upgrade-str).

  - Message-loss risk during broker rolling updates: During an upgrade, brokers restart one by one at different stages, temporarily reducing cluster availability; a broker failure during this period may increase the risk of message loss.
  - Ordered procedures for Kafka broker and client upgrades: Completing the procedures in the stated order allows brokers and clients to be upgraded without downtime.
  - Kubernetes 1.25 required for Strimzi 0.45.1
  - Cluster Operator first in the Kafka upgrade sequence
  - Kafka version and `spec.kafka.metadataVersion` updates for KRaft upgrades: The upgrade covers all Kafka brokers and client applications.
  - ZooKeeper-based Kafka upgrade prerequisites: Upgrading all Kafka brokers and client applications requires updating the Kafka version and `inter.broker.protocol.version`.
  - Upgrades and downgrades between KRaft-based clusters supported: This support is available from Strimzi 0.39.
  - Incremental upgrade from the previous minor version to Strimzi 0.45.1
  - Multi-version Strimzi upgrades across differing Kafka version support: If the new Strimzi version does not support the current Kafka version, it reports an unsupported-version error. The documented resolution is a Kafka version supported by the new Strimzi version, set through `spec.kafka.version`.
  - Upgrade path from Strimzi versions before 0.22: The documented path passes through 0.22, converts custom resources to `v1beta2`, and uses a version with `ControlPlaneListener` disabled (or enabled by default with the gate disabled) before enabling the feature gate and reaching 0.45.1. Alternatively, the 0.22 custom resources can be installed and converted without first upgrading to 0.22.
  - Kafka broker version is not validated against its image: The Cluster Operator cannot verify that the image contains a broker of the requested Kafka version. The note cautions that the image should correspond to that Kafka version.
  - Kubernetes upgrade path and supported-version guidance
  - Kafka availability during Kubernetes upgrades
  - High-availability upgrade prerequisites: Topics need a replication factor of at least 3, with `min.insync.replicas` set one below it; the example uses `replicas: 3` and `min.insync.replicas: 2`. In a highly available environment, the Cluster Operator maintains a minimum number of in-sync replicas during the upgrade to avoid downtime.
  - Cluster Operator upgrades use the initial deployment method
  - Version-specific upgrade documentation
  - Existing Cluster Operator deployment required for this procedure
  - Downloaded Strimzi 0.45.1 release artifacts required
  - Account permissions to create or manage `CustomResourceDefinition`, `ClusterRole`, and `RoleBinding` resources
  - Recording prior Cluster Operator configuration changes: The new version overwrites those changes.
  - Custom resource updates for Strimzi 0.45.1 configuration options
  - Cluster Operator installation-file namespace updates: The procedure says the new installation files should reflect the namespace where the Cluster Operator runs. It also provides RoleBinding YAML commands for Linux and macOS.
  - Use modified environment variables in the cluster operator Deployment: The procedure calls for editing `060-Deployment-strimzi-cluster-operator.yaml` to use the modified variables.
  - Role binding updates for multi-namespace Cluster Operator deployments
  - Cluster-wide role binding recreation for all-namespace watches: The procedure calls for recreating the bindings that grant cluster-wide access if needed.
  - Updated configuration deployed with the rest of the installation resources during Cluster Operator upgrades: The procedure waits for rolling updates to complete after deploying the updated configuration with the rest of the installation resources.
  - Error reported when upgrading from unsupported Kafka versions
  - Procedure for unsupported Kafka versions during operator upgrades: When an operator upgrade encounters a Kafka version it does not support, the procedure says to change `spec.kafka.version` to a supported version.
  - Kafka version upgrade can follow a Cluster Operator upgrade: When no unsupported-Kafka-version error is returned, the procedure allows the upgrade to continue.
  - OperatorHub.io upgrades use OLM to change the Strimzi operators’ update channel
  - Stable-channel automatic updates may miss pre-installation upgrade steps: Updates can arrive without a channel change. The note says automatic updates are intended only for version-specific channels.
  - Operator Lifecycle Manager documentation for upgrading Operators with OperatorHub
  - Manual CRD updates after Helm chart upgrades: The `helm upgrade` command does not upgrade the CRDs. The new CRDs are available from the GitHub releases page or the Helm chart’s `crd` subdirectory.
  - Kafka version compatibility for Cluster Operator upgrades: Upgrading the Cluster Operator in this situation produces an unsupported Kafka version error. Kafka must be upgraded by changing `spec.kafka.version`; the note also gives a `kubectl` command for checking the Kafka resource’s status conditions and explains its placeholders.
  - Apache Kafka documentation on KRaft-based upgrade support
  - Kafka configuration compatibility check before upgrading: Before upgrading, the Kafka resource must have no configuration options unsupported by the target Kafka version.
  - Kafka metadata version compatibility check: A configured value must be supported by the target Kafka version. If `spec.kafka.metadataVersion` is not configured, Strimzi updates it to the current default after the Kafka-version update.
  - Guide to keeping `metadataVersion` as a string
  - Kafka upgrade procedure keeps `metadataVersion` at its current Kafka-version default: The procedure sets `Kafka.spec.kafka.version` to the target version, upgrading brokers to the new binaries while brokers and controllers can communicate during the upgrade.
  - Kafka image for the new Kafka version: When `Kafka.spec.kafka.image` is defined, the procedure calls for updating it to an image with the new Kafka version.
  - Kafka Connect and MirrorMaker version properties in the upgrade procedure
  - Custom images need rebuilding with the latest Strimzi base image
  - Dockerfile updates for manually built Kafka Connect images: Their Dockerfiles need to point to the latest base image and build configuration.
  - Higher `metadataVersion` can block Kafka downgrades: If the new `metadataVersion` is higher than the version targeted for downgrade, the downgrade may be impossible. Keeping an older Kafka version can also have support and compatibility implications.
  - Kafka client updates and compatibility checks: As part of the Kafka upgrade, all Kafka client applications should be updated to the new client-binary version and checked for compatibility. Coordinating with the team managing those applications may be needed.
  - ZooKeeper-based Kafka cluster upgrade requirements: Upgrading these clusters requires updating both the Kafka version and the inter-broker protocol version.
  - ZooKeeper-to-KRaft migration guidance: Switching to KRaft mode must be done separately from the Kafka upgrade.
  - Kafka version and inter-broker protocol updates for ZooKeeper-managed upgrades: The Kafka version and compatible inter-broker protocol version are updated in the Kafka resource. The protocol is set cluster-wide through `Kafka.spec.kafka.config`.
  - Kafka message-format settings for upgrades: From Kafka 3.0.0, the message-format version is assumed to match `inter.broker.protocol.version`; the message-format settings can be removed when that protocol version is updated, or set to the target Kafka version otherwise. A topic’s default message format comes from the broker’s `log.message.format.version`, though the topic can also set its own value.
  - Kafka message format version properties deprecated: In this setup, `log.message.format.version` is ignored and need not be set.
  - Kafka message-format override migration guidance: Before Kafka 3.0, the overrides could accommodate older clients; without them, brokers convert older-format messages at significant performance cost. When removing the overrides during a broker upgrade, the note recommends upgrading consumers first and producers next; Apache Kafka Java clients have supported the latest message format since 0.11.
  - Requirements for upgrading Kafka: The Cluster Operator must be running, and the Kafka resource must not contain configuration options unsupported by the target Kafka version.
  - Kafka upgrade guidance for protocol and message format versions: When upgrading from Kafka 3.8.0 to 3.9.1, keep configured `inter.broker.protocol.version` and `log.message.format.version` at `3.8` before changing the Kafka version. If these settings are not configured, Strimzi updates them to the current defaults after the Kafka version changes.
  - `log.message.format.version` and `inter.broker.protocol.version` require string values: Otherwise, these values may be interpreted as floating-point numbers.
  - Kafka upgrades keep protocol settings at current-version defaults: For a 3.8.0-to-3.9.1 upgrade, both settings remain at `3.8` as brokers roll to the new binaries. Old and new binaries run at the same time during the rollout, and the unchanged inter-broker protocol lets them continue communicating.
  - Kafka downgrade incompatibility after an `inter.broker.protocol.version` change: After the upgrade, the downgraded cluster will not understand the resulting persistent-metadata schemas, including messages written to `__consumer_offsets`.
  - Latest Strimzi base image required for manually built custom images: The Dockerfile’s base image and build configuration need updating for the rebuild.
  - Inter-broker protocol version update guidance: For Kafka 3.9.1, the example sets `inter.broker.protocol.version` to `3.9` while leaving `log.message.format.version` at `3.8`.
  - Kafka upgrade version-setting prerequisite: The procedure updates `log.message.format.version` to the new version and waits for the Cluster Operator to update the cluster. For Kafka 3.9.1, the example sets both `log.message.format.version` and `inter.broker.protocol.version` to `3.9`.
  - FIPS procedure prerequisites
  - FIPS mode enabled for Kafka clusters: The procedure requires upgrading the Cluster Operator to 0.33 or newer while keeping `FIPS_MODE` disabled, and renewing cluster and clients CA certificates as needed: use `force-renew` for Cluster Operator-generated CAs, or update the CA secret and increase `ca-cert-generation` for your own CAs. SCRAM-SHA-512 users with passwords shorter than 32 characters need new passwords updated in clients; after the certificates and passwords meet the requirements, removing `FIPS_MODE` restarts the Cluster Operator and rolls all operands.

## Security (2)

### Jetty URI parsing and gzip buffer-release flaws addressed

- Affects: Strimzi installations using the bundled Jetty component (Inferred)
- [CVE-2024-13009](https://www.cve.org/CVERecord?id=CVE-2024-13009), [CVE-2024-6763](https://www.cve.org/CVERecord?id=CVE-2024-6763)

Before this release, `HttpURI` could validate an invalid URI authority differently from a browser, and a gzip error while inflating a request body could cause a buffer to be released incorrectly, corrupting or sharing data between requests. Strimzi now uses Jetty 9.4.57.v20241219, which the pull request says is intended to address both issues.

> * Bump jetty to address CVE-2024-6763 and CVE-2024-13009 by @fvaleri in https://github.com/strimzi/strimzi-kafka-operator/pull/11536

### `json-smart` stack-exhaustion flaw fixed

- Affects: Installations using the Strimzi cluster operator or Kafka 3.8.x/3.9.x third-party libraries
- Additional action: `LIMIT_JSON_DEPTH` for manually created `JSONParser` instances or custom parser options
- [CVE-2024-57699](https://www.cve.org/CVERecord?id=CVE-2024-57699)

With `json-smart` 2.5.0 through 2.5.1, specially crafted JSON containing many opening braces could exhaust the stack and cause denial of service. Strimzi now uses `json-smart` 2.5.2 in the cluster operator and Kafka 3.8.x/3.9.x third-party libraries, fixing this for `JSONParser` default modes such as `MODE_RFC4627`.

> * Added json-smart 2.5.2 as dependency to fix CVE-2024-57699 on 0.45.x release branch by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11411
> * Fixed CVE-2024-57699 by overriding json-smart dependency with 2.5.2 version.

## Breaking changes (3)

### Kafka 3.9.1 added as the supported default

- Affects: Kafka resources using the default version or configured to use Kafka 3.9.1 (Inferred)

Evidence: “This PR adds support for Kafka 3.9.1 to Strimzi.” https://github.com/strimzi/strimzi-kafka-operator/pull/11364

> * Add support for Kafka 3.9.1 by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11364

### Release candidates published directly to Sonatype

- Affects: Strimzi release-candidate artifacts consumed from Sonatype or OSS staging repositories (Inferred)

Previously, release candidates went through a staging repository; this release publishes them straight into Sonatype instead.

Evidence: “As there is no staging repository in new Sonatype portal, we decided to release RCs directly into the repository and not using any other staging repository.” https://github.com/strimzi/strimzi-kafka-operator/pull/11508

> * [release-0.45.x] Release RC artifacts directly to Sonatype by @im-konge in https://github.com/strimzi/strimzi-kafka-operator/pull/11508

### Topic Operator switches to unidirectional management

- Affects: Topic Operator setups upgrading from a Strimzi version that uses the Bidirectional Topic Operator
- Action required: After upgrading the Cluster Operator, delete the `KafkaTopic` resources managing `strimzi-store-topic` and `strimzi-topic-operator`. Retain `consumer-offsets` and `transaction-state` in Kafka, but annotate their `KafkaTopic` resources with `strimzi.io/managed="false"`; after reconciliation confirms they are no longer managed, delete those resources.

Previously, the Topic Operator could also perform operations on topics within the Kafka cluster; it now leaves independently managed topics alone.

> When deploying the Topic Operator to manage topics, the Cluster Operator enables unidirectional topic management.
> This means that the Topic Operator only manages Kafka topics associated with KafkaTopic resources and does not interfere with topics managed independently within the Kafka cluster.
> Previously, the Topic Operator worked in bidirectional mode, which meant it could also perform operations on topics within the Kafka cluster.

https://strimzi.io/docs/operators/0.45.1/deploying.html

## Important fixes (3)

### KRaft migration metadata loss after rollback fixed

- Affects: KRaft migrations

After a previous rollback, restarting a KRaft migration could cause cluster metadata loss because the `/migration` znode had not been deleted.

> * Fixed bug which may lead to metadata loss within the cluster when restarting a KRaft migration after a previous rollback due to missing `/migration` znode deletion.

### MirrorMaker 2 additional-volume handling fixed

- Affects: Strimzi setups using additional volumes in MirrorMaker 2 templates (Inferred)

Generated pods include configured ConfigMap and PVC volumes and their mounts.

> * Cherry-pick OPA and Additional Volumes in MM2 fixes to 0.45 release branch by @scholzj in https://github.com/strimzi/strimzi-kafka-operator/pull/11030

### KRaft migration rollback cleanup fixed

- Affects: clusters using KRaft migration (Inferred)

After a rollback, restarting KRaft migration without deleting the `/migration` znode could skip the initial ZooKeeper-to-KRaft sync and lead to metadata loss. Rollback cleanup now deletes that znode, resetting migration status so a later migration runs the initial sync.

> * Fixed missing deletion of the migration znode on KRaft migration rollback by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11271

## Enhancements (3)

- Kafka 3.9.1 support
- `--ignore-path` allowed for Kafka Connect Kaniko builds (To use: Specify `--ignore-path` in `additionalKanikoOptions`)
- Project names added to published Strimzi Maven POMs

## Dependency updates (8)

8 more, none mentioning security

- `* Dependency updates (Vert.x 4.5.13, Netty 4.1.118.Final).`
- `* Cherry-pick bumped Vert.x 4.5.12 and Netty 4.1.117.Final to 0.45.x release branch by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11079`
- `* Cherry-pick update Fabric8 Kubernetes Client to 7.1.0 to 0.45.x release branch by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11265`
- `* Chery-pick Update Vert.x to 4.5.13 to 0.45.x release branch by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11266`
- `* Chery-pick Update Fabric8 Kubernetes Client to 7.2.0 to 0.45.x release branch by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11395`
- `* [release-0.45.x] Bump test-containers to 0.109.2 by @im-konge in https://github.com/strimzi/strimzi-kafka-operator/pull/11547`
- `* [release-0.45.x] Bumped OAuth 0.15.1 by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11545`
- `* Updated bridge to 0.31.2 by @ppatierno in https://github.com/strimzi/strimzi-kafka-operator/pull/11564`

## Other changes (10)

- OPA system-test fixture updated
- Initial changelog section for 0.45.1
- Migration rollback documentation for controller node pool PVC deletion
- KRaft migration rollback cleanup documentation
- KRaft rollback and full migration system test added
- CodeQL workflow `actions/cache` plugin updated to v4.2.1
- Apache Kafka staging repository removed from the 0.45.x Maven POM
- Release 0.45.x artifact publishing moved to Sonatype Central
- Documentation on dedicated broker and controller nodes for production
- Documentation on combining dual-role and single-role nodes

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
