Manager does not detect Route CRD that appears after startup
維護者通常 1 天內回覆
@atishay-kasliwal 已經在處理了。
開始於 2026年11月19日。
評估
研究方向
Start with cmd/main.go and trace how util.IsRouteAPIFound() controls routev1.AddToScheme; then read the startup call to InspectCluster. The report says both checks happen only once, so investigate where API discovery or CRD events could update the scheme after startup. Done means a Route CRD created after manager startup can be reconciled without restarting the manager.
由索引模型根據 Issue 內容生成。
描述
Describe the bug
On a GKE management cluster where HyperShift is installed through GitOps, the OpenShift GitOps manager starts before the routes.route.openshift.io CRD exists. The manager registers the Route type only at startup: cmd/main.go calls routev1.AddToScheme only when util.IsRouteAPIFound() is true, and InspectCluster runs once. A Route API that appears later is never added to the scheme, so every Route reconciliation fails with:
no kind is registered for the type v1.Route in scheme "pkg/runtime/scheme.go:111"
The ArgoCD instance stays phase=Failed with Reconciled=False until the manager pod is restarted with the Route API present. This is not an edge case for this deployment shape: because the cluster installs HyperShift through Argo CD, the manager always starts before HyperShift creates the Route CRD.
To Reproduce
- Create a GKE cluster and install the OpenShift GitOps 1.22.2 operator, so the manager starts while
route.openshift.iois absent. - Confirm the API group is absent:
kubectl --context "$CTX" api-resources --api-group=route.openshift.io
Expected: no resources listed.
- Install HyperShift through Argo CD. The sync creates
routes.route.openshift.ioroughly 47 minutes after the manager started. - Wait for the operator to reconcile the
ArgoCDinstance.
Actual results
ArgoCD reports phase: Failed, Reconciled: False with the scheme error above. The operator log repeats the same error on each Route reconcile and it persists indefinitely.
Expected behavior
The manager should handle a Route CRD that appears after startup. Two approaches that would satisfy the use case:
- re-run API discovery periodically or on CRD watch events and add the Route type to the scheme once
routes.route.openshift.iois available, or - watch for the CRD itself instead of doing a one-time check at startup.
No claim is made about which approach upstream prefers.
Environment
- OpenShift GitOps 1.22.2, release 7 (
openshift-gitops-operator.v1.22.2) - Advisory RHBA-2026:83602 (2026-10-15)
- Bundle
registry.redhat.io/openshift-gitops-1/gitops-operator-bundle@sha256:2ba58c50f771c08bbc5a1d8572c7b058df55bc71667cc66026d42840eb4bee6d - Operator source
redhat-developer/gitops-operatoratcf32771491be2ed7f03d43d279171d6c265f2d2b - Downstream source
rh-gitops-midstream/releaseat219dafaaf0e9431f8b468b4932f337f3ce397058 - GKE
1.36.4-gke.2377000on Autopilot,REGULARchannel - Config Sync 1.23.3, Argo CD installed by Config Sync from a public bootstrap repo
Additional context
Timeline observed on 2026-11-13 (UTC):
- 16:00:03: manager pod started in
gitops-operator. - 16:46:57: HyperShift CRDs landed on the management cluster API server after the root ApplicationSet synced the management-cluster applications.
- 16:47:00:
routes.route.openshift.iocreated. - 16:47:11 to 17:08:
ArgoCDreportedphase=Failed,Reconciled=False; the scheme error repeated on every Route reconcile. - 17:08: manager pod deleted manually. The replacement registered the Route scheme and the instance reached
phase=Available,Reconciled=True. No Route object was created because every route switch in the instance spec is disabled.
Source reference at the released commit cf327714:
cmd/main.gocallsroutev1.AddToSchemeonly whenutil.IsRouteAPIFound()returns true.InspectClusterruns once at startup.
GITOPS-11466 (PR #1321, release backport #1354, commit 139db716ae65182624429cc1d78cd5a1c625f15c) makes Route detection independent of the Config API. That fixes a manager that starts after the Route CRD exists but with the Config API absent. It does not cover a Route API that appears after startup.
Workaround, either variant restores the manager without a code change:
- Install the Route CRD before the GitOps manager starts.
- Restart the manager once
routes.route.openshift.ioexists:
kubectl --context "$CTX" get crd routes.route.openshift.io
kubectl --context "$CTX" -n "$OPERATOR_NS" rollout restart deployment/openshift-gitops-operator-controller-manager
kubectl --context "$CTX" -n "$OPERATOR_NS" rollout status deployment/openshift-gitops-operator-controller-manager
This was found during a bounded development validation on a disposable GKE cluster. It is not a support or certification statement about this configuration.
- 主要語言
- Go
- 星號
- 190
- 分支
- 367
- 平均合併
- 4 天 2 小時
- 30 天內合併 PR
- 63
環境準備
- 提供 Dockerfile 或 Docker Compose 檔案
- 有 Pull Request 範本
- 閱讀貢獻指南
從這裡開始
- 先讀完整個 Issue,再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫,在一個分支上完成修改。
- 送出 Pull Request,並在描述裡引用這個 Issue 編號。
redhat-developer/gitops-operator 的其他 Issue
-
NamespaceManagement revocation checks labels on the wrong namespace可能已有人在做 @Rizwana777 於 3 天前認領。 未關閉kind/bug
難度 4/5 3-5 天 新手友好度 25/100
redhat-developer/gitops-operator#1559 · 1 則留言 · 已指派 1 人 ·
維護者通常 1 天內回覆
-
Deleting an ArgoCD instance removes RBAC belonging to same-named instances可能已有人在做 @Rizwana777 於 3 天前認領。 未關閉kind/bug
難度 3/5 1-2 天 新手友好度 45/100
redhat-developer/gitops-operator#1558 · 已指派 1 人 ·
維護者通常 1 天內回覆
-
NamespaceManagement cleanup can delete RBAC in unrelated namespaces可能已有人在做 @Rizwana777 於 3 天前認領。 未關閉kind/bug
難度 4/5 3-5 天 新手友好度 22/100
redhat-developer/gitops-operator#1557 · 已指派 1 人 ·
維護者通常 1 天內回覆
-
ApplicationSet tokenRef strict mode changes only apply after the controller is restarted可能已有人在做 @Gabrielandre02 於 5 天前認領。 未關閉
redhat-developer/gitops-operator#1554 · 1 則留言 · 已指派 1 人 ·
維護者通常 1 天內回覆
-
Enqueue the previous Argo CD instance when NamespaceManagement consent is withdrawn可能已有人在做 @Gabrielandre02 於 5 天前認領。 未關閉
難度 4/5 3-5 天 新手友好度 25/100
redhat-developer/gitops-operator#1530 · 1 則留言 ·
維護者通常 1 天內回覆
查看 redhat-developer/gitops-operator 的全部 Issue
相似的 Issue
-
benchmark
難度 1/5 1-3 小時 新手友好度 72/100
維護者通常 1 天內回覆
-
難度 2/5 1-3 小時 新手友好度 77/100
mozilla-ai/cq#601 ·
維護者通常 1 天內回覆
-
documentation migration
難度 2/5 1-3 小時 新手友好度 75/100
維護者通常 1 天內回覆
-
難度 2/5 1-3 小時 新手友好度 84/100
維護者通常 1 天內回覆
-
難度 2/5 1-3 小時 新手友好度 78/100
維護者通常 1 天內回覆