Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Hotfixes for Version 2.26

This section provides details on all hotfixes available for version 2.26. Hotfixes are critical updates released between our major and minor versions to address specific issues or vulnerabilities. These updates ensure the system remains secure, stable, and optimized without requiring a full version upgrade.

Version
Date (MM/DD/YYYY)
Internal ID
Description

2.26.58

11/04/2026

RUN-44340

Fixed an issue where renaming a role during the upgrade to 2.26 left orphaned AccessRule objects on the cluster that could not be deleted.

2.26.58

11/04/2026

RUN-41512

Fixed an issue where a resumed workload was shown as Completed instead of Running.

2.26.58

11/04/2026

RUN-44155

Fixed an issue where workload status notifications stopped being sent when a workload pod in the cluster was in a failed state.

2.26.58

11/04/2026

RUN-43419

Fixed an issue where a workload with sub-groups, such as master and worker, could be scheduled across different topology domains even when a required topology level was set.

2.26.58

11/04/2026

RUN-42660

Fixed an issue where distributed workloads submitted with a YAML file on MNNVL systems were not placed within a single NVL domain.

2.26.58

11/04/2026

RUN-44248

Fixed a security vulnerability related to GO-2026-6629 with severity HIGH.

2.26.58

11/04/2026

RUN-44238

Fixed a security vulnerability related to CVE-2026-84782 with severity HIGH.

2.26.58

11/04/2026

RUN-44186

Fixed a security vulnerability related to CVE-2026-111943 with severity HIGH.

2.26.58

11/04/2026

RUN-44159

Fixed an issue where distributed workload submission failed with a Bad Gateway error when the policy set defaults for only the worker or only the master.

2.26.58

11/04/2026

RUN-44153

Fixed an issue where the UI login flow did not use the recommended authentication settings.

2.26.58

11/04/2026

RUN-43897

Fixed an issue where the authentication client allowed direct password login.

2.26.58

11/04/2026

RUN-42865

Fixed security vulnerabilities related to CVE-2026-84304, CVE-2026-56855, and CVE-2026-78662 with severity HIGH.

2.26.58

11/04/2026

RUN-44030

Fixed a security vulnerability related to CVE-2026-84782 with severity HIGH.

2.26.58

11/04/2026

RUN-43982

Fixed a security vulnerability related to GHSA-qp49-qgx5-5m26 with severity HIGH.

2.26.58

11/04/2026

RUN-43967

Fixed a security vulnerability related to GHSA-hrxh-6v49-42gf with severity HIGH.

2.26.58

11/04/2026

RUN-43960

Fixed a security vulnerability related to CVE-2026-72897 with severity HIGH.

2.26.58

11/04/2026

RUN-43822

Fixed a security vulnerability related to CVE-2026-56859 with severity HIGH.

2.26.58

11/04/2026

RUN-43550

Fixed a security vulnerability related to CVE-2026-58213 with severity HIGH.

2.26.58

11/04/2026

RUN-43349

Fixed an issue where GPU fraction workloads failed to run on arm64 GPU nodes.

2.26.58

11/04/2026

RUN-43169

Fixed an issue where runai inference distributed port-forward failed with "workload's waiting time was exceeded" on distributed inference workloads.

2.26.58

11/04/2026

RUN-43096

Fixed an issue where distributed inference workloads from the NGC catalog failed to start when deployed as a guest.

2.26.58

11/04/2026

RUN-43140

Fixed an issue where runai workload describe returned "workload not found" for a newly submitted workload for up to about 25 seconds after creation.

2.26.58

11/04/2026

RUN-43040

Fixed an issue where the Stop action was available for inference workloads, which do not support stopping, leaving the workload stuck in the Stopping state.

2.26.58

11/04/2026

RUN-43249

Fixed an issue where the CLI replaced single quotes with double quotes in commands submitted after --.

2.26.58

11/04/2026

RUN-43180

Fixed an issue where the CLI did not apply an explicit --master-gpu-devices-request 0 to the master pod of distributed PyTorch workloads.

2.26.58

11/04/2026

RUN-43259

Fixed security vulnerabilities related to GO-2026-6303 and GHSA-hfg8-hc9c-6c3h with severity HIGH.

2.26.58

11/04/2026

RUN-43179

Fixed an issue where TFJob workloads failed to schedule after a cluster upgrade.

2.26.58

11/04/2026

RUN-43111

Fixed an issue where the idle GPU time limit scheduling rule did not pause workspaces.

2.26.58

11/04/2026

RUN-43089

Fixed an issue where the Workloads grid was slow to load and could time out.

2.26.58

11/04/2026

RUN-42356

Fixed an issue where inference workloads remained in a pending state after an OpenShift upgrade.

2.26.50

09/22/2026

RUN-43118

Fixed a security vulnerability related to CVE-2026-56853 with severity HIGH.

2.26.50

09/22/2026

RUN-43082

Fixed a security vulnerability related to GHSA-wg65-39gg-5wfj with severity HIGH.

2.26.50

09/22/2026

RUN-42905

Fixed an issue where copying a workspace with default UID/GID-source policy settings caused the copy to be rejected due to security identity values being resubmitted as explicit inputs.

2.26.50

09/22/2026

RUN-42289

Fixed an issue where the idle GPU time limit was not enforced for training workloads, allowing idle workloads to run past the configured limit.

2.26.50

09/22/2026

RUN-42727

Fixed an issue where the CLI rejected the --preferred-pod-topology-key value for NVL72 GPU cliques because it exceeded the 63-character label value limit.

2.26.50

09/22/2026

RUN-42663

Fixed an issue where the external endpoint option for LWS workloads was disabled and prompted for Knative configuration on clusters where Knative was not installed.

2.26.50

09/22/2026

RUN-42584

Fixed an issue where the UI displayed a success notification when an access rule creation was rejected by the API, instead of showing the error.

2.26.50

09/22/2026

RUN-42485

Fixed an issue where excessive error logs were generated after upgrade.

2.26.50

09/22/2026

RUN-42743

Fixed a security vulnerability related to CVE-2026-56852 with severity HIGH.

2.26.50

09/22/2026

RUN-42927

Fixed security vulnerabilities related to GHSA-3ccp-42pg-hgv6, GHSA-8rxv-jg7p-wvg3, and GHSA-fgjj-px3w-67xx with severity HIGH.

2.26.50

09/22/2026

RUN-42712

Fixed an issue where the researcher-service component remained deployed after being deprecated, due to a RunaiConfig toggle that was not updated.

2.26.50

09/22/2026

RUN-42935

Fixed an issue where a RayJob configured with MNNVL required failed before pod creation, due to an incorrect resource path in the default RayJob Karta.

2.26.50

09/22/2026

RUN-42664

Fixed an issue where uploading SAML IdP metadata for an existing SSO integration caused SSO login failures.

2.26.50

09/22/2026

RUN-42494

Fixed an issue where the Access Rules view for a user did not display inherited SSO group rules, showing only directly assigned rules.

2.26.50

09/22/2026

RUN-42632

Fixed an issue where the CLI installer reported successful shell configuration on unsupported shells, leaving the CLI unusable.

2.26.50

09/22/2026

RUN-41737

Fixed an issue where runai nodepool list reported incorrect GPU and node counts on clusters with more than 50 nodes.

2.26.50

09/22/2026

RUN-24516

Fixed an issue where telemetry responses for a disconnected cluster in the Node Pools view returned null instead of an empty array for GPU values.

2.26.50

09/22/2026

RUN-41758

Fixed an issue where, on very large environments, event sync from the cluster was rejected with an HTTP 413 (Request Entity Too Large) error.

2.26.50

09/22/2026

RUN-42484

Fixed an issue where a workload's Service was configured to target the OAuth proxy port even when the externalAuthUrlEnabled flag was set to false, causing connections to the workload to fail.

2.26.50

09/22/2026

RUN-42148

Fixed an issue where the UI and API allowed creating a shared Data Source from an existing PersistentVolumeClaim that was already used as a shared volume, an unsupported combination that caused continuous creation of orphaned PersistentVolumes. This combination is now blocked in both the UI and API.

2.26.50

09/22/2026

RUN-42486

Fixed a security vulnerability related to CVE-2026-60005 with severity HIGH.

2.26.50

09/22/2026

RUN-42366

Fixed a security vulnerability related to CVE-2026-39821 with severity HIGH.

2.26.50

09/22/2026

RUN-42328

Fixed a security vulnerability related to CVE-2026-62290 with severity HIGH.

2.26.50

09/22/2026

RUN-42327

Fixed a security vulnerability related to GHSA-hrxh-6v49-42gf with severity HIGH.

2.26.50

09/22/2026

RUN-41987

Fixed a security vulnerability related to GHSA-w879-237q-wc7r with severity HIGH.

2.26.50

09/22/2026

RUN-41760

Fixed an issue where cluster-sync could crash and restart during event processing on very large environments.

2.26.50

09/22/2026

RUN-42012

Fixed a security vulnerability related to CVE-2026-46600 with severity HIGH.

2.26.50

09/22/2026

RUN-42267

Fixed an issue where access keys did not inherit the user's UID, GID, and supplemental group claims.

2.26.50

09/22/2026

RUN-41991

Fixed a security vulnerability related to GHSA-f5wc-c3c7-36mc with severity CRITICAL.

2.26.50

09/22/2026

RUN-41898

Fixed a security vulnerability related to GHSA-jxpm-75mh-9fp7 with severity HIGH.

2.26.50

09/22/2026

RUN-41681

Fixed an issue where Data Volumes continuously created new PersistentVolumes when their corresponding PersistentVolumeClaim failed to be created first.

2.26.50

09/22/2026

RUN-42136

Fixed an issue where annotations added to a workload template were not propagated from the job to the pod when submitting workloads via the CLI.

2.26.50

09/22/2026

RUN-41855

Fixed an issue where inference endpoints exposed with --external-url did not enforce configured authentication.

2.26.50

09/22/2026

RUN-41791

Fixed an issue where the UI did not reflect an overridden preemptibility setting for inference workloads.

2.26.50

09/22/2026

RUN-36995

Fixed an issue where the ingress path-rewrite annotation used for inference workloads was incompatible with the official HAProxy ingress controller.

2.26.33

08/05/2026

RUN-41143

Fixed an issue where selecting a single node pool for a Project in the UI was not saved, causing the Project to inherit node pools from the parent Department instead.

2.26.27

07/30/2026

RUN-41720

Fixed an issue where inference throughput and latency metrics stopped updating cluster-wide when a Knative 1.19+ inference pod emitted more than one HTTP status code.

2.26.27

07/30/2026

RUN-41566

Fixed an issue where the device plugin failed to register with kubelet when devicePluginBindings was enabled, preventing GPU resources from being advertised to the cluster.

2.26.27

07/30/2026

RUN-41489

Fixed an issue where fractional GPU preloaders caused certain applications to crash on startup in fractional GPU workloads.

2.26.27

07/30/2026

RUN-41161

Fixed an issue where workloads with endpoint access set to "Specific users and service accounts" became stuck.

2.26.27

07/30/2026

RUN-41563

Fixed a security vulnerability (GHSA-xhf5-7wjv-pqxp, severity: High) in the application-controller service.

2.26.27

07/30/2026

RUN-41140

Fixed an issue in Nodes > Pod View where the pagination total included historical pods while the table displayed only active pods.

2.26.27

07/30/2026

RUN-41486

Fixed a security vulnerability related to CVE-2026-42530 with severity HIGH.

2.26.27

07/30/2026

RUN-41514

Fixed an issue where the tool connection URL returned 404 on Training and Workspace workloads when path-based routing was enabled.

2.26.27

07/30/2026

RUN-41445

Fixed an issue where the upgrade migration job failed with an ImagePullBackOff error on self-hosted installations.

2.26.27

07/30/2026

RUN-41378

Fixed an issue where email notification delivery stopped working after upgrading when the SMTP authentication type was set to "None".

2.26.27

07/30/2026

RUN-41210

Fixed an issue where workload status change email notifications were sent with blank content.

Last updated