Vendor682 problems· 14 reviewed

GitHub

682 incident problems in GitHub environments.

All problems (682)

CICD-1218A reusable deployment workflow succeeds but later cleanup deletes the artifactA reusable workflow promotes a build, but compliance review later finds the evidence package already deleted.CI/CDAdvanced20 minProCICD-536A reusable workflow signs container images correctlyA reusable workflow signs container images correctly focuses on ci-cd-release-safety and asks the reader to isolate Permission Denied in GitHub. 실무에서는 ci-cd-release-safety 문제를 볼 때 실패 단계만 보지 말고 최근 변경, 이미지 태그, 시크릿 주입, 롤백 가능 여부를 먼저 함께 확인하는 편이 빠릅니다.CI/CDIntermediate20 minProCICD-1233A self-hosted runner can clone the repo but package restore failsA self-hosted runner reaches GitHub and artifact storage but repeatedly fails on package restore against one internal or external registry.CI/CDAdvanced20 minProCICD-1221Reusable workflow secret is marked required but still arrives blankA deployment moved behind a reusable workflow and only the environment-backed secret fails to appear.CI/CDAdvanced20 minProCICD-1211Matrix jobs all succeed but the release step still sees only the last computed outputA matrix workflow computes per-target metadata, yet the downstream stage reads only one value because outputs are not aggregated the way the team assumed.CI/CDAdvanced21 minProCICD-028Retry policy re-runs destructive migration job twiceRetry policy re-runs destructive migration job twice is a hands-on troubleshooting drill. A generic retry wrapper helps flaky steps but becomes dangerous when applied to a migration or cleanup command. GitHub GitHub Actions Workflows needs to be checked by narrowing scope, rec...CI/CDIntermediate21 minProCICD-1212OIDC deploy works in the main workflow but fails after being moved into a reusable fileThe same shell commands now lose cloud auth because the reusable boundary no longer exposes the permission surface the original workflow relied on.CI/CDAdvanced22 minProCICD-1206Reusable workflow inherits secrets but one environment secret is still blank at runtimeThe reusable deployment starts normally, yet one secret resolves blank because the call boundary does not expose the environment-scoped secret the way the team assumed.CI/CDAdvanced22 minProCICD-1202Deploy key works for one checkout path but package restore still fails on a second private repositoryOne private repository is accessible, yet restore still breaks because a second dependency path needs a different credential scope than the first successful checkout.CI/CDAdvanced23 minProK8S-1185PodDisruptionBudget blocks node drainA public hardening guide recommended a conservative PDB. Later, replica counts dropped, but the old minAvailable still blocks every drain.KubernetesAdvanced23 minProCICD-1210Reusable workflow runs by tag but an internal checkout still resolves against caller contextThe reusable workflow is versioned by tag, yet one internal action path still follows caller context and breaks unexpectedly.CI/CDAdvanced23 minProCICD-1204Container registry login succeeds but build still pulls anonymouslyA registry login step copied from public examples looks green. The later builder still pulls anonymously because it uses a different execution context than the shell that performed the login.CI/CDAdvanced24 minProK8S-1200Controller pod runs but never actsA controller upgrade follows public guidance. The pod restarts cleanly, but reconciliations stop because the leader-election path cannot update its lease object.KubernetesAdvanced24 minProK8S-1205Controller starts fine after upgrade but CRD drift makes new objects fail validation on only one clusterThe controller image upgrade succeeds, but one cluster still rejects new objects because its CRD version or schema was not updated along with the controller.KubernetesAdvanced24 minProCICD-1192Docker layer cache grows until the self-hosted runner exhausts disk and silently poisons later jobsBuild jobs keep succeeding until the runner fills with old image layers and future runs fail in strange places because disk pressure is now shared across every pipeline.CI/CDAdvanced24 minProK8S-1195Draining one node silently breaks a local-path workload that was never truly portableA maintenance drain appears routine, but one workload fails badly because it depended on local storage behavior that was never made portable across nodes.KubernetesAdvanced24 minProCICD-1199Container image scan passes on build but release still ships an older vulnerable tagThe current build artifact is scanned successfully, yet the release pipeline promotes a different tag or digest that was never covered by the scan result.CI/CDAdvanced25 minProCICD-1189Matrix integration fan-out turns one shared preview stack into a false failure factoryParallel jobs hit the same preview environment and make stable application code look flaky under shared-state contention.CI/CDAdvanced25 minProCICD-054Buildx multi-arch pipeline failsThe image builds successfully for one architecture, but the overall publish stage fails when temporary layers and cache exports exceed the runner disk budget.CI/CDAdvanced26 minProCICD-1197OIDC deploy works for the workflow caller but fails inside a reusable deployment workflowThe top-level job can authenticate to AWS, yet the reusable deployment workflow fails because the assumed identity boundary differs inside the called workflow.CI/CDAdvanced26 minPro