AWS
339 incident problems in AWS environments.
먼저 읽을 가이드
추천 문제
All problems (339)
CICD-053OCI Helm chart publish succeeds but cluster still pulls the previous chartThe registry receives the new chart artifact, yet the deployment job keeps resolving the previous digest because the chart reference and version bump were not updated together.CI/CDIntermediate20 minProCICD-667A preview environment uses feature flags from production defaults and hides...A preview environment uses feature flags from production defaults and hides... focuses on ci-cd-workflow-debugging and asks the reader to isolate Permission Denied in AWS. 실무에서는 ci-cd-workflow-debugging 문제를 볼 때 실패 단계만 보지 말고 최근 변경, 이미지 태그, 시크릿 주입, 롤백 가능 여부를 먼저 함께 확인하는 편이 빠릅니다.CI/CDIntermediate21 minProCICD-055CodeBuild local cache reuses stale base layer after package repository changedBuild time looks great, but the produced image quietly carries an older package baseline because local Docker layer cache survives a repository-side base image update.CI/CDIntermediate21 minProK8S-1210CoreDNS stays degraded after node changesA new self-managed worker pool is added and CoreDNS remains degraded even though the managed pool looks healthy.KubernetesIntermediate22 minProCICD-726A preview environment uses feature flags from production defaults and hides...A preview environment uses feature flags from production defaults and hides... focuses on ci-cd-workflow-debugging and asks the reader to isolate Permission Denied in AWS. 실무에서는 ci-cd-workflow-debugging 문제를 볼 때 실패 단계만 보지 말고 최근 변경, 이미지 태그, 시크릿 주입, 롤백 가능 여부를 먼저 함께 확인하는 편이 빠릅니다.CI/CDIntermediate23 minProCICD-1609A Spinnaker artifact account fix lands and one pipeline still deploys the old objectOne Spinnaker pipeline still deploys the old object after account cleanup.CI/CDIntermediate8 minProCICD-1599A Spinnaker artifact binding is fixed and one pipeline still deploys the old packageOne Spinnaker pipeline still deploys the old package after artifact account cleanup.CI/CDIntermediate8 minProCICD-1579A Spinnaker webhook stage points to the new endpoint and one region still calls the old hostOne webhook stage keeps calling the old host after endpoint migration.CI/CDIntermediate8 minProCICD-1569A Spinnaker bake publishes the right manifest and one region still deploys the old base imageOne region deploys the old base image after an otherwise correct bake.CI/CDIntermediate9 minProCICD-1480A release automation bot closes the incident and users still fetch the old binaryMost users receive the new release immediately while one architecture continues downloading the previous binary.CI/CDIntermediate10 minProCICD-1470A release bot marks the artifact fresh while users still download the previous buildA release looks complete in chatops while a subset of downloads still resolves to the old object.CI/CDIntermediate10 minProK8S-1294A custom metrics HPA reads values successfully but never scalesA workload is migrated between namespaces and autoscaling appears frozen even though the metric backend is healthy.KubernetesIntermediate14 minProCICD-1335A Helm deploy from CodeBuild loses AWS auth only in one projectA private CodeBuild environment is hardened with VPC endpoints and later only one deploy job starts failing AWS auth inside the build container.CI/CDIntermediate14 minProK8S-1300A sidecar log shipper silently drops linesA logging change lands and one workload begins losing the first or last lines around every rotation event.KubernetesIntermediate14 minProK8S-1288An HPA sees CPU correctly but still never scalesAn application chart is refactored, CPU spikes are visible, and the HPA appears idle without errors.KubernetesIntermediate14 minProCICD-159A release pipeline reuses the same artifact name across two environments, and the staging package overwrites the production-ready build in shared storageThe objects are all present, but one naming collision turns an environment boundary into a storage race.CI/CDIntermediate16 minProCICD-174A multi-architecture promotion pipeline validates only one platform manifest during a failover rehearsalThe tag exists and one worker pool succeeds, yet another architecture never receives the complete promoted artifact. Normal traffic masked the issue until the standby or alternate path became active under rehearsal conditions.CI/CDIntermediate17 minProCICD-143The deployment job waits for image replication, but only the amd64 manifest arrives before the timeout and mixed-arch clusters fail progressivelyThe release gate checks one success condition, yet the fleet needs a broader replication contract than the workflow currently verifies.CI/CDIntermediate17 minProCICD-1678A reusable workflow widens token permissions centrally and one caller still deploys with a narrow scopeA reusable deployment workflow was fixed, but one repository still cannot assume the cloud role.CI/CDIntermediate11 minPro