Release Notes

v1.0.9

Changes

• The monitoring-and-autoscaling/alertmanager probe is now disabled by default. To use it, set alertmanagerProbe.enabled: true in the module configuration; the observability module is still required. The probe sends a synthetic test alert, and the silence it creates only covers the alertmanager of the observability module, so notification channels connected to Prometheus directly, such as an external Alertmanager, kept receiving the test alert.
• The extensions/dashboard probe is removed. It no longer appears in the probe list, the web UI and the availability API. Remove it from SLA reports and queries that referenced it; already recorded data stays in the database and expires with the regular retention.
• The module namespace is now checked against the restricted Pod Security Standards profile. Violations are reported as warnings only and do not block workloads. The exceptions the availability agent needs, such as host networking and its node-local data directory, are declared in a SecurityPolicyException resource.

Bug fixes

• The availability agent now starts on clusters where Pod Security Admission enforces the baseline profile by default, for example on Talos Linux.

Security

• Fixed a YAML injection through the external authentication URLs and the allowed source ranges in the status page and web UI settings. These values are now quoted when rendered, and source ranges must be valid CIDRs.
• Fixed CVE-2026-56855 and CVE-2026-78662 in the module images.

v1.0.8

Changes

• Fixed the module build for Deckhouse CSE.

v1.0.7

Security

• Fixed various vulnerabilities in third-party libraries and the Go standard library. Two findings that do not affect the module are documented as not exploitable.

v1.0.6

Bug fixes

• The module no longer fails to start on clusters without the virtualization module. Whether the image and the virtual machine class for the VM lifecycle probe are ready is now checked by the probe itself, which reports Unknown until they are.
• Fixed the detection of the default storage class for the VM lifecycle probe: the probe could be enabled on clusters where its disk could not be created.

v1.0.5

Features

• VM lifecycle probe. Upmeter now monitors the virtualization subsystem. Every 10 minutes the virtualization/vm-lifecycle probe creates a temporary virtual machine from the image set in virtualizationProbe.clusterVirtualImageName, live-migrates it to another node, attaches, grows and detaches an extra disk, and deletes everything it created. The probe stays disabled while the virtualization module is off, no image is configured or no storage class for disks is available. A missing or not yet ready image is reported as Unknown and does not affect the SLA.

v1.0.4

Bug fixes

• Fixed the node group probes on Deckhouse Virtualization Platform and other clouds with a single zone: nodes without a zone label no longer turn the probe result into Unknown.

v1.0.3

Bug fixes

• Fixed the module registry secret being created in the wrong namespace.

v1.0.2

Changes

• Module-specific registry credentials are now merged with the global registry credentials instead of replacing them.

v1.0.1

Security

• Fixed various vulnerabilities in third-party libraries.

v1.0.0

Changes

• First stable release of the upmeter module.

v0.2.2

Bug fixes

• Fixed the removal of the old upmeter StatefulSet during the migration to the standalone module.
• Fixed the registry secret name referenced by the module workloads.

v0.2.1

Changes

• The module now requires Deckhouse 1.76.0 or newer.

v0.2.0

Changes

• The module is now available in the Deckhouse SCE edition.

Enhancements

• Updated the web UI: fixed the mute button, tooltips, favicon and graph rendering.

v0.1.0

Features

• Initial release of the upmeter module as a standalone module of the Deckhouse Kubernetes Platform.