🐛 Kargo v1.10.7 is a patch release gathering three bug fixes, including a fix for Projects whose namespaces could become permanently stuck in Terminating.
🐛 Bug Fixes
-
Project namespaces no longer get stuck terminating (#6444): When a Project was deleted, Kubernetes garbage collection controllers (
namespace-controller,generic-garbage-collector,kube-controller-manager) attempt to delete the replicated secrets inside the terminating namespace. The replicated-resource-delete admission webhook was denying these requests, leaving namespaces permanently stuck inTerminating. The webhook now allows requests from well-known Kubernetes system controller identities, as well as from members of thesystem:mastersgroup for break-glass scenarios. -
Pipeline watch reconnects on tab switch instead of canceling (#6440): Switching away from and back to a browser tab tore down the pipeline Stages/Warehouses watch, which could leave the pipeline view showing stale data after the underlying stream was throttled, frozen, or silently dropped while hidden. The watch is now kept alive while the tab is inactive and reconnects when the tab becomes visible again, healing the stream. It is still cleaned up on unmount and when the project or warehouses change.
-
Spurious "Missing queryFn" error toast on the project page (#6430): On the project pipelines page, an error notification occasionally appeared reading
Missing queryFn: '["connect-query",...,"methodName":"QueryFreight",...]'. The freight watch was writing to aqueryFreightcache key that had no active query backing it, creating an orphaned entry that crashed on refetch. The watch now updates only caches with an active query (including warehouse-filtered variants), eliminating the confusing toast.
Full Changelog: v1.10.6...v1.10.7