Redpanda Helm Chart Release Notes

This page lists the changes in each Redpanda Helm chart 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 chart release changes the operator sidecar image that runs in the broker Pods, so upgrading the chart restarts all brokers in a rolling restart. The Upgrade notes column says whether a release has anything else to know before you upgrade.

If you deploy Redpanda with the Redpanda Operator, see Redpanda Operator Release Notes instead. To check which versions of Redpanda, Kubernetes, and Helm each release supports, see Kubernetes Compatibility.

Version Highlights Upgrade notes

v26.2.7
2026-11-20

The default Redpanda version installed by the chart is now v26.2.7, and the operator sidecar image tag is now v26.2.7.

Yes

v26.2.6
2026-11-11

Clients in the broker Pods now verify a listener that has a trustStore, a disabled external listener is no longer advertised, and the bundled Console subchart deploys Redpanda Console v3.14.0 by default.

Yes

v26.2.5
2026-10-23

A broker is no longer left in maintenance mode after a rolling restart, the default Redpanda version is now v26.2.6, and the configurator and fs-validator script volumes have shorter names, which restarts the brokers on upgrade.

Yes

v26.2.4
2026-09-24

Adds the commonAnnotations value, makes Redpanda v26.2.3 the default version, and removes the image.tag compatibility check.

Yes

v26.2.3
2026-09-03

Sidecar fixes stop superuser status from being briefly revoked from connected clients and keep users-file passwords out of the sidecar log, and the default Redpanda version is now v26.2.2.

Yes

v26.2.2
2026-08-17

The preStop lifecycle hook no longer stalls Pod termination or stops a broker that is still draining.

No

v26.2.1
2026-07-28

Adds beta Gateway API TLSRoute support for external access and opt-in host-level tuning, and fixes helm upgrade failures under Helm 4’s default server-side apply when SASL is enabled.

Yes

v26.2.7 (2026-11-20)

The default Redpanda version installed by the chart is now v26.2.7, and the operator sidecar image tag is now v26.2.7.

Before you upgrade
  • The default Redpanda version installed by the chart changes to v26.2.7.

Install or upgrade to this version
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda redpanda/redpanda \
  --namespace <namespace> \
  --create-namespace \
  --version 26.2.7 \
  --values <values-file>

This chart deploys Redpanda v26.2.7 by default. Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade Redpanda in Kubernetes.

Improvements

  • The default Redpanda version installed by the chart is now v26.2.7.

  • The operator sidecar image tag is now v26.2.7.

v26.2.6 (2026-11-11)

Clients in the broker Pods now verify a listener that has a trustStore, a disabled external listener is no longer advertised, and the bundled Console subchart deploys Redpanda Console v3.14.0 by default.

Before you upgrade
  • Brokers with a trustStore on an internal Admin API, Kafka API, or Schema Registry listener whose certificate also has a CA (caEnabled) get a new server-ca-bundle init container.

  • The bundled Console subchart now deploys Redpanda Console v3.14.0 by default.

Install or upgrade to this version
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda redpanda/redpanda \
  --namespace <namespace> \
  --create-namespace \
  --version 26.2.6 \
  --values <values-file>

This chart deploys Redpanda v26.2.6 by default. Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade Redpanda in Kubernetes.

Improvements

  • The bundled Console subchart now deploys Redpanda Console v3.14.0 by default.

  • The operator sidecar image tag is now v26.2.6.

Bug fixes

  • Clients in the broker Pods, including rpk, the sidecar, the lifecycle hooks, and the post-install Job, now verify an internal Admin API, Kafka API, or Schema Registry listener that has a trustStore, including when the listener’s certificate also has a CA (caEnabled). Brokers with a trustStore on one of these listeners restart once on upgrade.

  • An external Kafka or HTTP listener with enabled: false is no longer advertised. Previously, the chart advertised an address for a listener that Redpanda was never told to bind.

v26.2.5 (2026-10-23)

A broker is no longer left in maintenance mode after a rolling restart, the default Redpanda version is now v26.2.6, and the configurator and fs-validator script volumes have shorter names, which restarts the brokers on upgrade.

Before you upgrade
  • Upgrading an existing release changes the Pod template, so the brokers restart.

  • If you override the configurator or fs-validator volumes or volumeMounts through podTemplate, re-key the override from the old <fullname>-configurator or <fullname>-fs-validator name to configurator or fs-validator. An override keyed on the old name silently stops matching.

  • The default Redpanda version installed by the chart changes to v26.2.6.

Install or upgrade to this version
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda redpanda/redpanda \
  --namespace <namespace> \
  --create-namespace \
  --version 26.2.5 \
  --values <values-file>

This chart deploys Redpanda v26.2.6 by default. Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade Redpanda in Kubernetes.

Improvements

  • The configurator and fs-validator script volumes on the Redpanda StatefulSet are now named configurator and fs-validator rather than <fullname>-configurator and <fullname>-fs-validator, which keeps the configurator volume name within the Kubernetes 63-character limit for releases whose full name is 51 characters or longer. Upgrading restarts the brokers, and a podTemplate override keyed on an old volume name must be re-keyed to the new name.

  • values.schema.json is now 3-5x smaller, which reduces the time that every helm template, helm install, and helm upgrade spends compiling it. The schema accepts the same values as before, except that arrays nested inside a list or map of objects, such as podTemplate.spec.containers[].args, now accept null as well as a list.

  • The default Redpanda version installed by the chart is now v26.2.6, whose image no longer trusts the compromised eMudhra certificate authorities (CVE-2026-111943).

  • The operator sidecar image tag is now v26.2.5, which is built with Go 1.26.10 and rejects certificates that chain to the compromised eMudhra certificate authorities (CVE-2026-111943).

Bug fixes

  • config.rpk.tune_*: false is honored again when tuning.apply_host_tuners is enabled. Previously, every default-enabled tuner ran regardless of the opt-out.

  • A broker is no longer left in maintenance mode after a rolling restart when its admin API takes longer than half the termination grace period (45s by default) to start. Previously, that broker held no partition leadership and prevented the next broker in the roll from being drained before its restart.

v26.2.4 (2026-09-24)

Adds the commonAnnotations value, makes Redpanda v26.2.3 the default version, and removes the image.tag compatibility check.

Before you upgrade
  • The default Redpanda version installed by the chart is now v26.2.3.

  • The chart no longer checks image.tag for compatibility, so a Redpanda version older than the chart supports fails at runtime instead of at render time.

  • The force value no longer has any effect.

Install or upgrade to this version
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda redpanda/redpanda \
  --namespace <namespace> \
  --create-namespace \
  --version 26.2.4 \
  --values <values-file>

This chart deploys Redpanda v26.2.3 by default. Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade Redpanda in Kubernetes.

New features

  • The commonAnnotations value adds the given annotations to every object the chart renders, including the StatefulSet, for example argocd.argoproj.io/sync-wave: "1". The StatefulSet’s Pod template and volume claim templates are left unannotated, and annotations set on an individual resource take precedence.

Improvements

  • The operator sidecar image tag is now v26.2.4.

  • The default Redpanda version installed by the chart is now v26.2.3.

Bug fixes

  • resources.memory.container no longer shows the memory locking description in the Helm reference. That description belongs to resources.memory.enable_memory_locking.

  • helm install prints the post-install notes again.

Removed

  • The chart no longer checks image.tag for compatibility. Deploying a Redpanda version older than the chart supports now fails at runtime instead of with a render-time error, and the force value no longer has any effect.

v26.2.3 (2026-09-03)

Sidecar fixes stop superuser status from being briefly revoked from connected clients and keep users-file passwords out of the sidecar log, and the default Redpanda version is now v26.2.2.

Before you upgrade
  • The default Redpanda version installed by the chart changes to v26.2.2.

  • The operator sidecar image tag changes to v26.2.3.

Install or upgrade to this version
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda redpanda/redpanda \
  --namespace <namespace> \
  --create-namespace \
  --version 26.2.3 \
  --values <values-file>

This chart deploys Redpanda v26.2.2 by default. Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade Redpanda in Kubernetes.

Improvements

  • The operator sidecar image tag is now v26.2.3.

  • The default Redpanda version installed by the chart is now v26.2.2.

Bug fixes

  • The sidecar no longer momentarily revokes superuser status from connected clients when it starts or when a users secret is updated. This issue affected only Helm chart deployments, not clusters deployed by the operator.

  • The sidecar now writes the superusers cluster property only when the list changes, so sidecar restarts no longer bump the cluster config version. When a users file can’t be read, the sidecar retries instead of writing a partial list.

  • Malformed lines in a users file, which can contain passwords, are no longer written to the sidecar log.

v26.2.2 (2026-08-17)

The preStop lifecycle hook no longer stalls Pod termination or stops a broker that is still draining.

Install or upgrade to this version
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda redpanda/redpanda \
  --namespace <namespace> \
  --create-namespace \
  --version 26.2.2 \
  --values <values-file>

This chart deploys Redpanda v26.2.1 by default. Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade Redpanda in Kubernetes.

Improvements

  • The operator sidecar image tag is now v26.2.2.

Bug fixes

  • The preStop lifecycle hook no longer stalls Pod termination or sends SIGTERM to a broker that is still draining. The hook now skips the drain for a broker whose node ID is no longer a cluster member, and every Admin API call it makes has a timeout.

v26.2.1 (2026-07-28)

Adds beta Gateway API TLSRoute support for external access and opt-in host-level tuning, and fixes helm upgrade failures under Helm 4’s default server-side apply when SASL is enabled.

Before you upgrade
  • If you run the PVCUnbinder sidecar (statefulset.sideCars.pvcUnbinder.enabled), its ClusterRole gains read-only, cluster-scoped nodes list and storageclasses get. If you set rbac.enabled: false, add these to your out-of-band role.

Install or upgrade to this version
helm repo add redpanda https://charts.redpanda.com
helm repo update
helm upgrade --install redpanda redpanda/redpanda \
  --namespace <namespace> \
  --create-namespace \
  --version 26.2.1 \
  --values <values-file>

This chart deploys Redpanda v26.2.1 by default. Replace <values-file> with the file that holds your overrides. For the full procedure, see Upgrade Redpanda in Kubernetes.

New features

  • External listeners can use Gateway API TLSRoute for SNI-based routing through a Gateway that you manage (beta). To opt in, set type: tlsroute on an individual external listener, which overrides the cluster-wide external.type.

    Details

    The chart values under listeners.<proto>.external.<name>.type and external.gateway may still change in backwards-incompatible ways before GA. Production use is not recommended.

  • Set tuning.apply_host_tuners to give the tuning init container real access to the host, so that tuners that previously failed inside the Pod sandbox, such as disk_irq, disk_scheduler, and net, apply. This is opt-in (default false) and needs a privileged init container and hostPath volumes.

    Details

    Requires PSA privileged namespaces, or an OpenShift SCC that permits hostPath and privileged containers. tune_aio_events must remain enabled. Run at most one Redpanda Pod per node. In this release, config.rpk.tune_*: false does not turn off a tuner. Upgrade to v26.2.5 or later to turn off an individual tuner. Rollback is not automatic: tuning that is already applied persists until the node reboots, and the fstrim systemd timer installed on the host keeps firing until you remove it.

  • statefulset.persistentVolumeClaimRetentionPolicy sets the StatefulSet’s PersistentVolumeClaim retention policy. PVCs are retained by default. Set whenScaled: Delete to delete a broker’s PVC when it is removed by a scale-down, or whenDeleted: Delete to delete all PVCs when the StatefulSet is deleted.

Improvements

  • The bundled Console subchart is now version 3.9.0.

  • The PVCUnbinder sidecar’s ClusterRole now also grants read-only, cluster-scoped nodes list and storageclasses get, which the unbinder’s new pvc-rebinding-gate exemption uses. If you set rbac.enabled: false, add these to your out-of-band role, or the unbinder falls back to its conservative deferral.

    Details

    To turn off only the exemption and keep the PVCUnbinder running, set statefulset.sideCars.pvcUnbinder.disableStuckClaimExemption.

Bug fixes

  • Post-upgrade processing of the cluster configuration no longer produces configuration errors for properties that were removed across Redpanda versions.

  • Configuration job Pods no longer have labels that overlap with those of the StatefulSet Pods, so selectors meant for the StatefulSet Pods no longer match them.

  • Ballast file tuning now works. Previously, the ballast file did not reach the main container.

  • The bootstrap user password is now loaded from an externally managed Secret referenced by auth.sasl.bootstrapUser.secretKeyRef. Previously, SCRAM authentication failed from the operator and other chart-owned clients.

  • The ServiceMonitor now renders scheme: http or scheme: https in lowercase, as chart 25.3.x and earlier did. Previously, older prometheus-operator CRDs rejected the uppercase value with spec.endpoints[0].scheme: Unsupported value.

  • When a pre-stop or post-start lifecycle hook times out, the Pod logs now show a line containing "TIMEOUT after Ns" with the hook name and exit code. Previously, a timeout looked the same as a clean run.

  • A boolean cluster property set to false under config.cluster, such as cloud_storage_enable_remote_read: false, is no longer silently overwritten by tiered storage defaults.

  • helm upgrade no longer fails under Helm 4’s default server-side apply when SASL is enabled. Previously, the API server rejected the *-bootstrap-user Secret with metadata.managedFields must be nil, and later server-side rollbacks failed the same way.

  • The postStart lifecycle hook now clears maintenance mode left behind by a previous incarnation of the Pod that re-registered under a different node ID, for example after losing its data directory when running without persistent storage. Previously, the old node ID stayed in maintenance mode forever, which blocked every other broker from entering maintenance mode.