This update corrects the following known issues:
Corrected DockerHub authentication issue. The
withDockerRegistry step failed to authenticate with DockerHub in some scenarios. Now, the
withDockerRegistry step now uses the Docker login CLI when to authenticate with DockerHub.
Corrected authentication issue when not using the default Jenkins superuser. When the Authorize Project plugin was installed and configured to run builds as a user other than the default Jenkins superuser, the
withDockerServer steps were unable to find the specified credentials and so the steps failed. Users with
Credentials/UseItem permission (implied by Job/Configure permission) on a Pipeline Job can use credentials which that job has access to in the
Dockerfiles containing a FROM that used ARGs defined with default values did not work correctly with the docker DSL. Now, ARGs with default values are now handled correctly by the docker DSL.
Environment variables are now properly masked for pipeline steps. Pipeline steps that used the Jenkins Launcher API inside of a
withDockerContainer step did not mask environment variables. Now, all environment variables passed to Docker exec are now masked inside of the
A script for initializing Logstash and Topbeat is now available, which allows Elasticsearch to run in workers or controllers.
Kube Agent Management Plugin upgraded from
Sign up information from the Setup Wizard is sent to Segment.
Credentials masking works even when the password has special characters but does not include a new line. For example, if your secret is an SSH key which takes multiple lines, they might not be detected.
The license expiration date in the
about.md file was set to the JVM locale.
The expiration date is now displayed using the UTC format.
Jira Plugin had an increasing number of file leaks due to a non-close of
httpclient, which resulted in a "Too Many Files Open" error. Now, the plugin uses a single
httpclient instance to avoid selector number increase.
The BitBucket and GitHub Branch Source plugin "Checkout over SSH" falls back to using HTTPS when configured to use key on the agent. Now, the "Checkout over SSH" properly accepts and uses key on agent when configured.
When retry is used as a top-level option in a Declarative Pipeline and a stage fails, the retry attempts will skip all stages due to the original failure. With this fix, the original failure will be reset when the retry begins, allowing stages to execute normally during the retry attempt.
Users were unable to resume Pipelines that come with nested parallels. Parallel and load steps did not clean up internal state correctly after their executions completed, leading to failures upon resuming a Pipeline and memory leaks in some cases. The internal state of parallel and load steps is now cleaned up correctly when the step completes.
Values for choice parameters can be passed via the build step even if those values are not among the allowed values on the target job. The build step will fail with an error if an invalid choice value is specified.
Node.js plugin updated to allow installation of Artifactory plugin versions later than 3.0.0.
Elasticsearch clusters can fail due to excessive snapshot backup accumulation. A fix is available that limits the number of snapshots.
A bug has been fixed where a device could not be unmounted due to device locks from open file descriptors or a crashed/hung process, causing manual forced restart of workers.
The Update Center used to not show the open source Azure CLI plugin due to a conflicting artifactId. For new versions of CloudBees Products, the open source Azure CLI plugin is available in the Update Center. CloudBees Azure CLI plugin has been deprecated and should not be used.
When two plugins have the same ID (but different groupIds), the plugin information provided in Plugins Management did not belong to the proper one and updates for the wrong plugin were offered.
Information for both plugins can now be visualized when managing plugins.
An update to the
dna command fixes a crash when displaying terminal output with unspecified dimensions.
A bug has been fixed where Elasticsearch was mounted on build workers, which is unnecessary as Elasticsearch does not run on build workers.