Redpanda Operator Release Notes
This page lists the changes in each Redpanda Operator release in this release line, newest first. Each release shows what to know before you upgrade and the command to upgrade to that version.
Every operator release changes the sidecar image that runs in the broker Pods, so upgrading the operator restarts the brokers of Redpanda and StretchCluster resources.
For a summary of the major features in this release line, see What’s New in the Redpanda Operator. To check which versions of Redpanda, Kubernetes, and Helm each release supports, see Kubernetes Compatibility.
| Version | Highlights | Upgrade notes |
|---|---|---|
v26.1.15 |
Operator upgrades now reach broker Pods, a ShadowLink that has been failed over no longer reports |
|
v26.1.14 |
Operator upgrades no longer restart every broker when only the embedded Redpanda chart’s version changes, and TLS verification is fixed for clusters that use |
|
v26.1.13 |
NodePool brokers inherit an unset |
|
v26.1.12 |
|
|
v26.1.11 |
Rolling restarts wait for each broker’s Schema Registry to catch up before the next broker restarts, and fixes correct node pool labels and restore User, Role, and Group reconciliation on clusters with the Schema Registry listener disabled. |
|
v26.1.10 |
The default Redpanda version that the operator installs moves to v26.1.14. |
|
v26.1.9 |
Adds cross-namespace |
No |
v26.1.8 |
The bundled Redpanda Console version moves to v3.8.0. |
|
v26.1.7 |
Values in |
No |
v26.1.6 |
The operator Helm chart gains a |
No |
v26.1.5 |
The broker decommissioner works when |
|
v26.1.4 |
Rolling restarts no longer risk taking two brokers down at once, Redpanda and Console ServiceMonitors work again with older |
No |
v26.1.3 |
The operator no longer reconciles healthy Redpanda clusters in a tight loop, and it authenticates with SCRAM when the bootstrap user’s Secret is managed externally. |
No |
v26.1.2 |
Cluster configuration no longer fails after an upgrade because of properties removed between Redpanda versions, and migrating from Helm to the operator clears the leftover Helm field manager. |
|
v26.1.1 |
Adds group-based access control (GBAC), Schema Registry ACLs in every CRD that sets ACLs, and support for Kubernetes versions 1.32.x through 1.35.x. |
v26.1.15 (2026-11-20)
Operator upgrades now reach broker Pods, a ShadowLink that has been failed over no longer reports Synced=False, and the default Redpanda image tag is now v26.1.21.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.15 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
Bug fixes
-
When a broker rejects a cluster configuration write, the Redpanda resource’s
ConfigurationAppliedcondition and a Warning event now carry the broker’s per-property reason instead of a bareBad Request. -
Operator upgrades now reach broker Pods. The first upgrade to this release rolls brokers once if their Pod template changed since it was last applied.
Details
With
image,statefulset.sideCars.image, andstatefulset.initContainerImagepinned andlogging.usageStats.enabled: false, brokers roll only when a release changes the broker Pod spec. -
A ShadowLink whose link has been failed over no longer reports
Synced=Falseand re-sends the refused update on every reconcile. The operator stops updating a link that isfailing over,failover faulted, orfailed over, and reports those states instatus.state.Details
Such a link no longer reads the source cluster’s resource or Secrets, so deleting them after the failover doesn’t fail the sync. If the current spec was not applied before the failover started,
SyncedisFalsewith reasonConfigurationFrozen, without retrying. The newstatus.appliedGenerationrecords the generation of the spec last applied to the link. -
The V1 Cluster controller no longer removes the
cluster_idcluster configuration property once a minute. Previously, Redpanda wrote the property back each time, so the cluster configuration version climbed by two each minute.
v26.1.14 (2026-11-11)
Operator upgrades no longer restart every broker when only the embedded Redpanda chart’s version changes, and TLS verification is fixed for clusters that use issuerRef certificates or a listener trustStore.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.14 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
Improvements
-
The default Redpanda Console image for
Consoleresources is now v3.14.0. -
The default Redpanda image tag (
--redpanda-tag) is now v26.1.20.
Bug fixes
-
StretchCluster broker startup and liveness probes now set
timeoutSeconds: 5, so three Admin API responses slower than 1s no longer restart a healthy broker. Applying this fix restarts StretchCluster brokers once, and atimeoutSecondsset through a NodePoolpodTemplatestill overrides the new value. -
Upgrading the operator no longer restarts every broker just because the embedded Redpanda chart’s version changed. Brokers still restart when their Pod spec changes, such as a new sidecar image or, with
logging.usageStats.enabled, the operator version reported inREDPANDA_METRICS_K8S_CHART_VERSION. The upgrade to this release restarts brokers managed byRedpandaresources once to drop thehelm.sh/chartlabel from broker Pods. -
Topic resources without
spec.partitionsorspec.replicationFactorno longer fail every reconcile after the topic is created. The operator now leaves the topic’s current partition count and replication factor alone when those fields are unset. -
Resources that reference a Redpanda cluster through
cluster.clusterRef, such asUser,Topic,Schema, andRole, no longer fail when the cluster’s certificate setstls.certs.<cert>.issuerRef. ForissuerRefandsecretRefcertificates alike, the operator now trusts every certificate in the issued Secret’s key, not only the first, so a CA bundle or a full chain works. -
Clients in Redpanda and StretchCluster broker Pods (
rpk, the sidecar, the lifecycle hooks, and, for Redpanda resources, the post-install Job) now verify an internal listener that has atrustStore. Previously, the sidecar and the post-install Job could fail to connect to the Admin API, and thepostStarthook could retry forever when the listener’s certificate also had a CA and the trust store held only client CAs. When the certificate also has a CA (caEnabled), a newserver-ca-bundleinit container writes a bundle ofca.crtand the trust store, and the sidecar rewrites it when either changes. Brokers with atrustStoreon an internal Admin API, Kafka API, or Schema Registry listener restart once on upgrade.
v26.1.13 (2026-10-23)
NodePool brokers inherit an unset image from the Redpanda resource, the operator keeps the order of string-array cluster configuration values, and the operator and the default Redpanda image no longer trust the compromised eMudhra certificate authorities (CVE-2026-111943).
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.13 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
Improvements
-
values.schema.jsonis 3-5x smaller, sohelm template,helm install, andhelm upgradespend less time compiling it. The schema accepts the same values as before, except that arrays nested inside a list or map of objects, such aspodTemplate.spec.containers[].args, now acceptnullas well as a list. -
The default Redpanda image tag (
--redpanda-tag) is now v26.1.19. That image no longer trusts the compromised eMudhra certificate authorities (CVE-2026-111943).
Bug fixes
-
The operator no longer sorts string-array cluster configuration values, so order-sensitive properties such as
kafka_mtls_principal_mapping_rules,sasl_kerberos_principal_mapping, andiceberg_default_catalog_namespacekeep the order given in the Redpanda resource. Previously, aDEFAULTrule could sort ahead of custom rules and map every principal to its full DN. -
A
Consolethat references a Redpanda cluster throughclusterRefnow connects to the cluster’s NodePool brokers. Previously, once a NodePool migration drained the cluster’s own StatefulSet tostatefulset.replicas: 0, Console crash-looped withyou must specify at least one broker to connect to.Details
With
--enable-v2-nodepools, Console includes every NodePool bound to the cluster and re-renders when a NodePool is added, removed, or resized. -
Breaking change: NodePool brokers now inherit any unset
imagefield from the Redpanda resource instead of the operator’s default Redpanda image, which produced a mixed-version cluster when the Redpanda resource pinnedimage.tag. On upgrade, NodePools that leaveimageunset roll to the Redpanda resource’s image, so first setimage.tagon the Redpanda resource (or on the NodePool) to the version the NodePool runs today.Details
If the Redpanda resource’s image is older than the version a NodePool currently runs, the roll is a downgrade that Redpanda may refuse to start on.
-
The operator is now built with Go 1.26.10, so it rejects certificates that chain to the compromised eMudhra intermediate certificate authorities (CVE-2026-111943). The build also includes fixes for several HTTP/2 denial-of-service vulnerabilities.
-
golang.org/x/netis updated to v0.60.0,golang.org/x/cryptoto v0.57.0, and OpenTelemetry to v1.46.0 (logs v0.22.0) to address reported vulnerabilities, including a busy-spin in the OpenTelemetry log batch processor (CVE-2026-81872).
v26.1.12 (2026-09-24)
podTemplate now applies to the chart’s hook Jobs as well as the operator Deployment, cluster configuration is applied even when the post-install job is disabled, and Console 3.9 no longer crash-loops on clusters without SASL.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.12 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
New features
-
podTemplatenow applies to every workload the chart renders: the operator Deployment, the CRD installation Job, and the post-upgrade migration Job. Both Jobs now inherit its default PodsecurityContext(runAsUser: 65532), which previously reached only the Deployment.Details
To override one workload, use
deployment.podTemplate,jobs.crd.podTemplate, orjobs.migration.podTemplate. Each one layers on top ofpodTemplate, which makes the containersecurityContextof the hook Jobs configurable.Pod templates are merged with a form of strategic merge patch that keys on unique names as defined by Kubernetes. To override the settings of one container, specify the name of that container.
Bug fixes
-
Setting
post_install_job.enabled=falseno longer prevents cluster configuration from being applied. Cluster configuration is now derived from chart values instead of the post-install job, so environment injected throughpost_install_job.podTemplateno longer feeds it. Useconfig.extraClusterConfigurationorstorage.tiered.credentialsSecretRefinstead. -
The Console configuration rendered for a Redpanda resource no longer contains
kafka.sasl.enabled: trueandkafka.sasl.impersonateUser: trueunless the legacyconsole.consoleblock configures a login. Previously, Console 3.9 crash-looped on clusters without SASL, because it refuses to start with impersonation enabled while its own authentication is disabled.
v26.1.11 (2026-09-03)
Rolling restarts wait for each broker’s Schema Registry to catch up before the next broker restarts, and fixes correct node pool labels and restore User, Role, and Group reconciliation on clusters with the Schema Registry listener disabled.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.11 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
New features
-
Rolling restarts now wait for each broker’s Schema Registry to catch up before the next broker restarts, so a cluster keeps a consistent Schema Registry endpoint during an upgrade. The wait applies to Cluster, Redpanda, and StretchCluster resources, is skipped on clusters without a Schema Registry listener, and can be disabled with
--wait-for-schema-registry-sync=false. -
The single-cluster operator now supports
--leader-election-lease-duration,--leader-election-renew-deadline, and--leader-election-retry-periodfor tuning leader election. The defaults remain 15s, 10s, and 2s.
Bug fixes
-
Each node pool’s Pods now carry their own
cluster.redpanda.com/nodepoollabel. Previously, every pool after the first got the first pool’s label. -
The operator chart’s ServiceMonitor now sets
scheme: httpsin lowercase. Previously, olderprometheus-operatorCRDs rejected the uppercase value withspec.endpoints[0].scheme: Unsupported value. -
The sidecar no longer writes malformed users-file lines, which can contain passwords, to its log.
-
User, Role, and Group resources now reconcile on clusters with
listeners.schemaRegistry.enabled=false. Previously, the disabled listener blocked all ACL syncs.
v26.1.10 (2026-07-29)
The default Redpanda version that the operator installs moves to v26.1.14.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.10 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
v26.1.9 (2026-07-14)
Adds cross-namespace clusterRef references and client tuning options for ShadowLink resources, and fixes the User CRD so that it accepts a password set with value alone.
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.9 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
New features
-
clusterRefhas an optionalnamespacefield, whichShadowLink,Topic,User,Role,Schema, andGroupresources can use to reference a cluster in another namespace. When the field is not set, the cluster is resolved in the referencing object’s namespace.Details
A
ShadowLinkcan have its source and shadow clusters in separate namespaces on the same Kubernetes cluster. This topology is intended mainly for development and test clusters, where you can rehearse shadow-link replication and failover on a single Kubernetes cluster. For production disaster recovery, place the source and shadow clusters in separate Kubernetes clusters and failure domains. -
ShadowLink resources now expose
spec.clientOptionsfor tuning the Kafka client that the shadow cluster uses to fetch from the source cluster. Each field keeps the broker default when unset or set to 0, so existing ShadowLink resources are unaffected.Details
The fields are
fetchMinBytes,fetchWaitMaxMs,fetchMaxBytes,fetchPartitionMaxBytes,metadataMaxAgeMs,connectionTimeoutMs, andretryBackoffMs. They primarily trade replication latency against fetch batching efficiency.
v26.1.8 (2026-07-06)
The bundled Redpanda Console version moves to v3.8.0.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.8 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
v26.1.7 (2026-06-30)
Values in additionalCmdFlags now override the operator’s default command-line flags, and User deletion no longer gets stuck on clusters without an enterprise license.
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.7 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
Bug fixes
-
Values in
additionalCmdFlagsnow override the operator’s default command-line flags in the Deployment. Previously, a flag that the chart sets a default for, such as--enable-console, was silently dropped in favor of the default. -
Deleting a
Userno longer gets stuck on clusters without an enterprise license. Previously, when the user had no Schema Registry ACLs to delete, the cleanup failed with "Invalid license: not present" and the deletion never completed.
v26.1.6 (2026-06-17)
The operator Helm chart gains a priorityClassName value, clusterSpec.statefulset probes and volume mounts now reach the broker container, and configured sync intervals take effect for the Topic, User, and other resource controllers.
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.6 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
New features
-
The operator Helm chart has a top-level
priorityClassNamevalue, which you can set to a higher-priority PriorityClass so that the Redpanda Operator Pod is less likely to be preempted when the cluster is under resource pressure. The default is"", which uses the cluster’s global default PriorityClass, if any.
Bug fixes
-
The Topic controller no longer issues a no-op
IncrementalAlterConfigson every reconcile. It writes only the configuration keys that changed. -
Configured sync intervals, such as the Topic
spec.interval, now take effect for the Topic, User, Group, Schema, Role, and ShadowLink controllers. Previously, every status-only update of the referenced cluster resource re-enqueued all of these resources. -
clusterSpec.statefulset.extraVolumeMounts,livenessProbe, andstartupProbeare no longer silently dropped from theredpanda(broker) container. Previously, they applied only to thesidecarcontainer. -
clusterSpec.statefulsetoverrides (probes andextraVolumeMounts) are no longer silently dropped whenpodTemplate.spec.containersorpodTemplate.spec.initContainerscontains duplicate entries with the same name. The overrides now apply to the last duplicate, which is the entry the chart renders.
v26.1.5 (2026-06-04)
The broker decommissioner works when fullnameOverride is set.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.5 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
v26.1.4 (2026-05-21)
Rolling restarts no longer risk taking two brokers down at once, Redpanda and Console ServiceMonitors work again with older prometheus-operator CRDs, and the operator Helm chart can set a scrape interval and custom labels on its ServiceMonitor.
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.4 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
New features
-
The operator Helm chart’s
monitoringconfiguration has two new fields,scrapeIntervalandlabels.scrapeIntervalsets the scrape interval on the ServiceMonitor endpoint and is empty by default, which keeps the Prometheus global default, andlabelsadds custom labels to the ServiceMonitor metadata.
Bug fixes
-
The in-use features of an enterprise cluster now have a deterministic sort order. Previously, the changing order caused status changes that could lead to reconciliation storms.
-
The ServiceMonitor that the operator renders for a Redpanda cluster or a Console is rendered with a lowercase scheme,
scheme: httporscheme: https, as in chart 25.3.x and earlier. Olderprometheus-operatorCRDs rejected the uppercase value withspec.endpoints[0].scheme: Unsupported value. The operator Helm chart’s own ServiceMonitor still usesscheme: HTTPSin this release. -
A rolling restart no longer risks making two broker Pods unavailable at the same time. The operator now waits until every Pod on the latest StatefulSet revision is running and ready, and no Pod is terminating, before it restarts the next one.
v26.1.3 (2026-04-24)
The operator no longer reconciles healthy Redpanda clusters in a tight loop, and it authenticates with SCRAM when the bootstrap user’s Secret is managed externally.
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.3 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
Bug fixes
-
Ballast file tuning now propagates the ballast file to the main container. Previously, the tuning container did not mount the directory this needs.
-
The operator removes more historic field managers that caused problems with changes to StatefulSets.
-
The operator no longer re-enqueues a healthy Redpanda cluster about every 2 seconds. It settles into the 3-minute periodic requeue interval.
-
The operator no longer fails to authenticate to the Redpanda cluster with SCRAM when
bootstrapUser.secretKeyRefpoints at an externally managed Secret.
v26.1.2 (2026-04-03)
Cluster configuration no longer fails after an upgrade because of properties removed between Redpanda versions, and migrating from Helm to the operator clears the leftover Helm field manager.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.2 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
v26.1.1 (2026-04-01)
Adds group-based access control (GBAC), Schema Registry ACLs in every CRD that sets ACLs, and support for Kubernetes versions 1.32.x through 1.35.x.
|
Before you upgrade
|
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda-controller redpanda/operator \
--namespace <namespace> \
--create-namespace \
--version 26.1.1 \
--set crds.enabled=true \
--values <values-file>
Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade the Redpanda Operator.
New features
-
The RedpandaRole spec has an
Internalboolean field. Use it to manage Redpanda internal roles (prefixed with "__") with standard Kubernetes resource names. -
The operator supports group-based access control (GBAC), both as a standalone Group CRD and through Role-based principal binding.
-
All CRDs that support setting ACLs support Schema Registry ACLs.
-
The operator supports Kubernetes versions 1.32.x through 1.35.x.
-
The redpanda, redpanda-configurator, and sidecar containers set
allowPrivilegeEscalation: falseandrunAsNonRoot: truein their security contexts. This addresses GKE security findings. -
Controllers can now run in namespace-scoped mode with noise filtering.