Skip to content

Commit 9a4caba

Browse files
yankawaiyankawai
committed
docs(cluster-api): explain the omitted migration priority by the pinned API
CAPK v0.1.10 pins kubevirt.io/api v1.5.0, whose VirtualMachineInstanceMigrationSpec carries only vmiName; upstream #392 sets priority against v1.9.0. The typed migration does not compile against the pinned API, which is why the patch builds it unstructured. Assisted-by: LLM Signed-off-by: Yan Bondarenko <[email protected]>
1 parent 2e2b475 commit 9a4caba

1 file changed

Lines changed: 1 addition & 1 deletion

File tree

  • packages/system/capi-providers-infraprovider/images/cluster-api-provider-kubevirt

‎packages/system/capi-providers-infraprovider/images/cluster-api-provider-kubevirt/UPSTREAM.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
The build stays on CAPK v0.1.10, commit `6dbc11c4d0cbfbdcd618566538c10b0d4a07516b`, and preserves its runtime image and API dependencies. The patch backports [guest node lookup #374](https://github.com/kubernetes-sigs/cluster-api-provider-kubevirt/pull/374), [uncordon after restart #389](https://github.com/kubernetes-sigs/cluster-api-provider-kubevirt/pull/389), and [migration before drain/delete #392](https://github.com/kubernetes-sigs/cluster-api-provider-kubevirt/pull/392). It bounds submission waiting to ten minutes, separately from the active migration and drain fallback windows, preserves each deadline across reconciles, and omits migration priority, which KubeVirt reserves for virt-controller even with MigrationPriorityQueue enabled.
1+
The build stays on CAPK v0.1.10, commit `6dbc11c4d0cbfbdcd618566538c10b0d4a07516b`, and preserves its runtime image and API dependencies. The patch backports [guest node lookup #374](https://github.com/kubernetes-sigs/cluster-api-provider-kubevirt/pull/374), [uncordon after restart #389](https://github.com/kubernetes-sigs/cluster-api-provider-kubevirt/pull/389), and [migration before drain/delete #392](https://github.com/kubernetes-sigs/cluster-api-provider-kubevirt/pull/392). It bounds submission waiting to ten minutes, separately from the active migration and drain fallback windows, preserves each deadline across reconciles, and omits migration priority. CAPK v0.1.10 pins `kubevirt.io/api` v1.5.0, whose `VirtualMachineInstanceMigrationSpec` carries only `vmiName`; #392 sets `priority` against v1.9.0, where the field exists. The typed migration from #392 does not compile against the pinned API, which is why the patch builds the migration as an unstructured object and leaves priority to KubeVirt.
22

33
`make image` runs the backported regression tests, builds the manager, and writes the resulting tag and digest to `images/cluster-api-provider-kubevirt.tag`. The InfrastructureProvider reads that file through `spec.deployment.containers[].imageUrl`, supported by CAPI operator v0.19.0. The compressed provider manifest contains the additional VMIM create permission. Keeping the image reference out of the gzip lets the existing image collection, promotion, and nightly mirror code see and rewrite it.
44

0 commit comments

Comments
 (0)