The following problems are commonly reported after a CloudBees CI upgrade, each with a short description and a link to an article describing the potential solution. Use the list as a quick reference if an upgrade has already failed, and you’re trying to find the cause.
Common issues
The following common issues are the most frequent causes of failed or degraded upgrades reported to CloudBees Support:
-
Outdated plugins outside the CloudBees Assurance Program (CAP) block the dependency cascade on restart, or controller-only plugins installed on the operations center block its restart. For more information, refer to Upgrade plugins with Beekeeper Upgrade Assistant.
-
A Configuration as Code (CasC) bundle or its plugin catalog can fail validation against the target envelope. For more information, refer to Determine plugin compatibility using CasC.
-
The JDKs used by the controller, operations center, and agents must all be upgraded to the same version. For more information, refer to CloudBees CI on modern cloud platforms: Migrate to Java 21 or CloudBees CI on traditional platforms: Migrate to Java 21.
-
High Availability (HA) nodes must never run mismatched versions. For more information, refer to High Availability (active/active).
-
A newer Helm chart may require a higher Kubernetes version, or introduce newly mandatory values. For more information, refer to Supported platforms for CloudBees CI on modern cloud platforms and Controllers can’t use shared Kubernetes cloud after changing Agents.SeparateNamespace.Enabled.
-
GitOps or CloudBees CI deployer service accounts may lack cluster-scoped Role-Based Access Control the chart needs. For more information, refer to Troubleshooting deployer service account RBAC permission errors during a Helm upgrade on OpenShift.
-
Gateway API
HTTPRouteresources can fail to attach even with correct namespace labels. For more information, refer to Verify Kubernetes Gateway API prerequisites. -
A shared-agent leasing race condition on pre-2.528.2.34846 instances can strand jobs in the queue, and can worsen during an upgrade window. For more information, refer to Shared agents stuck in unknown state and jobs queue indefinitely after upgrade.
-
Introducing JobConfigHistory retention limits for the first time on a large or old history store can delay controller startup by 25+ minutes. For more information, refer to JobConfigHistory Plugin Best Practices.
-
A customized
jgroups.xmlreferencing the removedCHMOD_FILE_PINGprotocol will fail to boot past 2.528.1.29783. For more information, refer to HA controller fails to start with "JGRP000002: unable to load protocol CHMOD_FILE_PING" after upgrading to 2.528.1.29783 or later. -
Custom Update Center certificates signed with SHA-1 are rejected after migrating to Java 21. For more information, refer to Custom Update Center certificate fails signature verification with a disabled SHA-1 algorithm error on RHEL 9 managed controllers.
-
Your instance is unavailable, unusable, or performing poorly right after an upgrade. For more information, refer to CloudBees Performance Decision Tree for troubleshooting.
-
After an instance upgrade, the builds are running slower than usual. For more information, refer to Troubleshoot Slow Builds.
-
HTML reports are failing after an instance upgrade. For more information, refer to What is Content Security Policy and how does it impact Jenkins?