Changelog v1.77.9
Fixes
- [cloud-provider-azure] Disable the node-ipam controller in the cloud-controller-manager to prevent duplicate pod CIDRs and conflicting routes.
- [cloud-provider-openstack] Fix OpenStackCluster rendering when network lists are missing or empty
- [cloud-provider-vcd] optimize VM cache refresh and avoid blocking cache reads
- [cloud-provider-vcd] preserve VM not-found errors to allow orphaned Node cleanup
- [cloud-provider-yandex] quote label values in generated MachineClass manifests
- [common] Update nginx in common/nginx-static to 1.30.7 to fix CVEs.
Pods using common/nginx-static (registry-proxy, basic-auth, deckhouse-tools, documentation) will be restarted. - [deckhouse-controller] Releases of a still-embedded module no longer raise manual approval alerts once the module has a deployed release.
- [node-manager] Disable HTTP keep-alives in kubernetes-api-proxy
/readyzprobes so connections close immediately after each check.
Static pods and DaemonSet pods ofkubernetes-api-proxywill restart. - [node-manager] Fix
StaticInstancegetting stuck in a bootstrap loop after cleanup, and status patches failing on legacy conditions. - [node-manager] Nodes trust the registry CA from a ModuleSource that has
dockerCfgset again.
bashible-apiserver no longer skips such ModuleSources, so their CA is added to the node configuration. Bashible will be rerun on all nodes. - [node-manager] Restore the default maxSurge behavior for NodeGroups with maxSurgePerZone set to 0.
- [registry-packages-proxy] Reject CSE node packages with untrusted manifest signatures, including requests through registry aliases and cached downloads.
The registry-packages-proxy pods will be restarted.
Chore
- [registrypackages] Update delivery-kit-sdk to v1.4.2 and remove C build dependencies from d8-bcheck.