GitLab 17.0 released with generally available CI/CD Catalog and AI Impact analytics dashboard
GitLab 17.0 released with generally available CI/CD Catalog, AI Impact analytics dashboard, hosted runners on Linux Arm, deployment detail pages, and much more!
These are just a few highlights from the 60+ improvements in this release. Read on to check out all of the great updates below.
To the wider GitLab community, thank you for the unbelievable 344 contributions you provided to GitLab 17.0!
At GitLab, everyone can contribute and we couldn't have done it without you!
Niklas van Schrick now has the hat trick with three MVPs and has become one of GitLab’s most consistent contributors with at least one merge request per milestone since GitLab 14.3.
Niklas was nominated by Magdalena Frankiewicz, Product Manager at GitLab, for contributing a feature to create custom webhook payload templates and then following it up with the ability to specify custom webhook headers.
“This solved a highly demanded 7-year-old feature request with 65 upvotes,” says Magdalena.
“Users can now fully design custom webhooks!”
Niklas is a member of the GitLab Core Team and helps the wider community and GitLab live up to our mission to enable everyone to contribute.
“During my journey, I interacted with a lot of different reviewers, maintainers, designers, technical writers, product managers, and probably more,” Niklas says.
“Everyone was helpful and did their best to help move issues and merge requests forward.”
Gerardo Navarro has been contributing to GitLab for over a year and takes home a second GitLab MVP award.
Gerardo was nominated for creating ongoing contributions towards a feature to show protected packages in the package registry list. This feature is part of a series of contributions related to the protected packages epic that intends to increase security by enabling fine-grained permissions to create, update, and delete packages from the package registry.
Many thanks to Gerardo Navarro and the rest of the team from Siemens for helping co-create GitLab.
“Thank you very much for appreciating our work with such a cool award,” says Gerardo.
“I feel honored. I am still learning a lot with every contribution.”
The CI/CD Catalog is now generally available. As part of this release, we’re also making CI/CD components and inputs generally available.
With the CI/CD Catalog, you gain access to a vast array of components created by the community and industry experts.
Whether you’re seeking solutions for continuous integration, deployment pipelines, or automation tasks, you’ll find a diverse selection of components tailored to suit your requirements.
You can read more about the Catalog and its features in the following blog post.
You’re invited to contribute CI/CD components to the Catalog and help expand this new and growing part of GitLab.com!
AI Impact is a dashboard available in the Value Streams Dashboard that helps organizations understand the impact of GitLab Duo on their productivity.
This new month-over-month metric view compares the AI Usage trends with SDLC metrics like lead time, cycle time, DORA, and vulnerabilities. Software leaders can use the AI Impact dashboard to measure how much time is saved in their end-to-end workstream, while staying focused on business outcomes rather than developer activity.
In this first release, the AI usage is measured as the monthly Code Suggestions usage rate, and is calculated as the number of monthly unique Code Suggestions users divided by total monthly unique contributors.
The AI Impact dashboard is available to users on the Ultimate tier for a limited time. Afterwards, a GitLab Duo Enterprise license will be required to use the dashboard.
We are excited to introduce hosted runners on Linux Arm for GitLab.com.
The now available medium and large Arm machine types, equipped with 4 and 8 vCPUs respectively, and fully integrated with GitLab CI/CD, will allow you to build and test your application faster and more cost-efficient than ever before.
We are determined to provide the industry’s fastest CI/CD build speed and look forward to seeing teams achieve even shorter feedback cycles and ultimately deliver software faster.
You can now link directly to a deployment in GitLab. Previously, if you were collaborating on a deployment, you had to look up the deployment from the deployment list. Because of the number of deployments listed, finding the correct deployment was difficult and prone to error.
From 17.0, GitLab offers a deployment details view that you can link to directly. In this first version, the deployment details page offers an overview of the deployment job and the possibility to approve, reject, or comment on a deployment in a continuous delivery setting. We are looking into further avenues to enhance the deployment details page, including by linking to it from the related pipeline job. We would love to hear your feedback in issue 450700.
GitLab Duo Chat just got a lot better. It now uses Anthropic Claude 3 Sonnet as the base model, replacing Claude 2.1 for answering most questions.
At GitLab, we apply a test-driven approach when choosing the best model for a set of tasks and authoring well-performing prompts. With recent adjustments to the chat prompts, we have achieved significant improvements in the correctness, comprehensiveness, and readability of chat answers based on Claude 3 Sonnet compared to the previous chat version built on Claude 2.1. Hence, we have now switched to this new model version.
A popular capability of GitLab Duo Chat is answering questions about how to use GitLab. While Chat offers various other capabilities, this particular functionality was previously only available on GitLab.com. With this release, we’re making it accessible to GitLab self-managed deployments as well, aligning with our commitment to delivering a delightful experience across all types of deployments.
Whether you’re a newcomer or an expert, you can ask Chat for help with queries like “How do I change my password in GitLab?” or “How do I connect a Kubernetes cluster to GitLab?”. Chat aims to provide helpful information to solve your problems more efficiently.
We enhanced the Value Streams Dashboard with an Overview panel. This new visualization addresses the need for executive-level insights into software delivery performance, and gives a clear picture of GitLab usage in the context of software development life cycle (SDLC).
The Overview panel displays metrics for the group level, such as number of (sub)groups, projects, users, issues, merge requests, and pipelines.
Introduced in GitLab 15.9, the CI/CD job token allowlist prevents unauthorized access from other projects to your project. Previously, you could allow access at the project level from other specific projects only, with a maximum limit of 200 total projects.
In GitLab 17.0, you can now add groups to a project’s CI/CD job token allowlist. The maximum limit of 200 now applies to both projects and groups, meaning a project allowlist can now have up to 200 projects and groups authorized for access. This improvement makes it easier to add large numbers of projects associated with a group.
The rules:exists CI/CD keyword has default behaviors that vary based on where the keyword is defined, which can make it harder to use with more complex pipelines. When defined in a job, rules:exists searches for specified files in the project running the pipeline. However, when defined in an include section, rules:exists searches for specified files in the project hosting the configuration file containing the include section. If configuration is split over multiple files and projects, it can be hard to know which exact project will be searched for defined files.
In this release, we have introduced project and ref subkeys to rules:exists, providing you a way to explicitly control the search context for this keyword. These new subkeys help you ensure accurate rule evaluation by precisely specifying the search context, mitigating inconsistencies, and enhancing clarity in your pipeline rule definitions.
You can now view the status of configuration changes made to your GitLab Dedicated instance infrastructure using the Switchboard configuration page.
All users with access to view or edit your tenant in Switchboard will be able to view changes in the Configuration Change log and track their progress as they are applied to your instance.
Currently, the Switchboard configuration page and change log are available for changes like managing access to your instance by adding an IP to the allowlist or configuring your instance’s SAML settings.
We will be extending this functionality to enable self-serve updates for additional configurations in coming quarters.
With this release, you can use the REST API to enable viewing Jira issues in GitLab. You can also specify one or more Jira projects to view issues from.
Until now, imported items were not easily identifiable. With this release, we’ve added visual indicators to items imported with direct transfer, where the creator is identified as a specific user:
For larger repositories, you can now view issues from multiple Jira projects in GitLab when you set up the Jira issue integration. With this release, you can:
Enter up to 100 Jira project keys separated by commas.
Leave Jira project keys blank to include all available keys.
When you view Jira issues in GitLab, you can filter the issues by project.
We’ve updated what happens when you delete an epic to better safeguard your project’s structure and data. It’s all about giving you more control and peace of mind while managing your projects.
Now, when you delete a parent epic, instead of deleting all its child records automatically, we preserve them by detaching the parent relationship first. This change provides you with a safer way to manage your epics, ensuring accidental deletions don’t result in losing valuable information.
We’ve improved issue boards to offer you clearer insights into your project’s timeline and phases. Now, with milestone and iteration details directly visible on issue cards, you can easily track progress and adjust your team’s workload on the fly. This enhancement is designed to make your planning and execution more efficient, keeping you in the loop and ahead of schedule.
You can now customize Value Streams Dashboard panels using a simplified schema-driven customizable UI framework. In the new format, the fields provide more flexibility of displaying the data and laying out the dashboard panels. With the new framework, administrators can track changes to the dashboard over time. This version history can help you revert to previous versions and compare changes between dashboard versions.
Using this customization, decision-makers can focus on the most relevant information for their business, while teams can better organize and display key DevSecOps metrics.
You can now integrate 1Password secrets management with the GitLab Duo plugin for JetBrains.
Developers can replace their personal access tokens in their JetBrains IDE settings with 1Password secrets references. This simplifies managing secrets, and enables seamless secrets rotation without manual token updates.
Previously, web commits and automated commits made by GitLab could not be signed. Now you can configure your self-managed instance with a signing key, a committer name, and email address to sign web and automated commits.
The after_script CI/CD keyword is used to run additional commands after the main script section of a job. This is often used for cleaning up environments or other resources that were used by the job. However, after_script commands did not run if a job was canceled.
As of GitLab 17.0, after_script commands will always run when a job is canceled. To opt out, see the documentation.
When using a CI/CD Catalog component, you might want to have it automatically use the latest version. For example, you don’t want to have to manually monitor all the components you use and manually switch to the next version every time there is a minor update or security patch. But using ~latest is also a bit risky because minor version updates could have undesired behavior changes, and major version updates have a higher risk of breaking changes.
With this release, you can opt to use the latest major or minor version of a CI/CD component. For example, specify 2 for the component version, and you’ll get all updates for that major version, like 2.1.1, 2.1.2, 2.2.0, but not 3.0.0. Specify 2.1 and you’ll only get patch updates for that minor version, like 2.1.1, 2.1.2, but not 2.2.0.
You can use the GitLab package registry to publish and download packages. Sometimes, packages fail to upload due to an error. Previously, there was no way to quickly view packages that failed to upload. This made it challenging to get a holistic view of your organization’s package registry.
Now you can filter the package registry UI for packages that failed to upload. This improvement makes it easier to investigate and resolve any issues you encounter.
From GitLab 17.0, you can install GitLab in FIPS mode with the agent for Kubernetes components enabled. Now, FIPS-compliant users can benefit from all the Kubernetes integrations with GitLab.
Sometimes there is more than one person involved in resolving a support ticket or
the requester wants to keep colleagues up-to date on the state of the ticket.
Now you can have a maximum of 10 external participants without a GitLab account on a
Service Desk ticket and regular issues.
External participants receive Service Desk notification emails for each public comment
on the ticket, and their replies will appear as comments in the GitLab UI.
Simply use the quick actions /add_email
and remove_email
to add or remove external participants with a few keystrokes.
DAST 5 supports both arm64 and amd64 architectures by default. This enables customers to choose the Runner host architecture and optimize cost savings.
Users of Dependency Scanning can now scan Android projects. To configure Android scanning, use the CI/CD Catalog component. Android scanning is also supported for users of the CI/CD template.
We resolved a Secret Detection bug that impacted remote rulesets. It’s now possible to override or disable rules via remote rulesets. Remote rulesets offer a scalable way to configure rules in a single place, which can be applied across multiple projects.
If you add a secondary email address to your user profile and do not verify it, that email address is now automatically deleted after three days. Previously, these email addresses were in a reserved state and could not be released without manual intervention. This automatic deletion reduces administrator overhead and prevents users from reserving email addresses that they do not have ownership of.
Previously, you could not edit an existing custom role and its permissions. Now, you can edit a custom role and its permissions without having to re-create the role to make a change.
Previously, setting up default branch protection options did not allow for the same level of configuration that the settings for protected branches did.
In this release, we have updated the default branch protection settings to provide the same experience that you have with protected branches.
This allows more flexibility in protecting your default branch and simplifies the process to match what already exists in the protected branch settings.
With the release of these custom permissions, you can reduce the number of Owners needed in a group by creating a custom role with these Owner-equivalent permissions. Custom roles allow you to define granular roles that give a user only the permissions they need to do their jobs, and reduce unnecessary privilege escalation.
Compliance operates on a sliding scale for many organizations as they strike a balance between meeting requirements and ensuring developer velocity is not impacted. Merge request approval policies help to operationalize security and compliance in the heart of the DevSecOps workflow - the merge request. We’re introducing a new fail open option for merge request approval policies to offer flexibility to teams who want to ease the transition to policy enforcement as they roll out controls in their organization.
When a merge request approval policy is configured to fail open, MRs will now only be blocked if a policy rule is violated and if that project has the security analyzer properly configured. If an analyzer is not enabled for a project or if the analyzer does not successfully produce results, the policy will no longer consider this a violation for the given rule and analyzer. This approach allows for progressive rollout of policies as teams work to ensure proper scan execution and enforcement.
The old implementation of the Vulnerability Report filters wasn’t scalable.
We were limited by horizontal space on the page. You can now use the filtered
search component to filter the Vulnerability Report by any combination of
status, severity, tool, or activity. This change allows us to add new filters,
like this proposed filter by identifier.
Previously, when a public group or project invited a private group, the private group was listed only in the Groups tab of the Members page, and private members were not visible to members of the public group. To enable better collaboration between members of these groups, we are now also listing all invited group members in the Members tab, including members from private invited groups. The source of membership will be masked from members that do not have access to the private group. However, the source of membership will be visible to users who have at least the Maintainer role in the project or Owner role in the group, so that they can manage members in their project or group. If the current user viewing the Members tab is unauthenticated or not a member of the group or project, they will not see the private group members. We hope this change will make it easier for group and project members to understand at a glance who has access to a group or project.
When importing projects from export files with many items of the same type (for example, merge requests or pipelines), sometimes some of those items weren’t imported.
In this release, we added an API endpoint that re-imports a named relation, skipping items that have already been imported. The API requires both:
A project export archive.
A type (issues, merge requests, pipelines, or milestones).
GitLab is expanding collaboration by updating our permissions. Now, users with the Reporter role can access Design Management features, enabling product teams to engage more directly in the design process. This change simplifies workflows and accelerates innovation by inviting broader participation from across your organization.
We reduced the minimum role required to relate issues and tasks from Reporter to Guest, giving you more flexibility to organize work across your GitLab instance while maintaining permissions.
We added a new metric to the Value Streams Dashboard: median time to merge. In GitLab, this metric represents the median time between when a merge request was created and when it was merged. This new metric measures DevOps health by identifying the efficiency and productivity of your merge request and code review processes.
By analyzing how this metric evolves in the context of other SDLC metrics, teams can identify low or high productivity months, understand the impact of new DevOps practices on the development speed and delivery process, reduce their overall lead time, and increase the velocity of their software delivery.
We expanded the epic sorting options available in the Roadmap view, providing you more flexibility in organizing and prioritizing your projects. You can now sort epics by created date, last updated date, and title. This enhancement lays the groundwork for even more advanced sorting capabilities in the future to help you manage epics more dynamically.
Opening Duo Chat directly from your editor in JetBrains is now even easier.
Use the default Alt+D keyboard shortcut (or set your own) to open Duo Chat quickly and type your question. Use the same keyboard shortcut to close the window.
Across an organization, it can be helpful to have the same templated response in issues, epics, and merge requests. These responses might include standard questions that need to be answered, responses to common problems, or good structure for merge request review comments. Project-level comment templates give you an additional way to scope the availability of templates, bringing organizations more control and flexibility in sharing these across users.
To create a comment template, go to any comment box on GitLab and select Insert comment template > Manage project comment templates. After you create a comment template, it’s available for all project members. Select the Insert comment template icon while making a comment, and your saved response will be applied.
We’re really excited about this iteration of comment templates and if you have any feedback, please leave it in issue 451520.
We’re also releasing GitLab Runner 17.0 today! GitLab Runner is the lightweight, highly-scalable agent that runs your CI/CD jobs and sends the results back to a GitLab instance. GitLab Runner works in conjunction with GitLab CI/CD, the open-source continuous integration service included with GitLab.
We have been hard at work on CI/CD components, including making the process of releasing components to the CI/CD Catalog a consistent experience. As part of that work, we’ve made releasing versions from a CI/CD job with the release keyword and the release-cli image the only method. All improvements to the release process will apply to this method only. To avoid breaking changes introduced by this restriction, make sure you always use the latest version of the image (release-cli:latest) or at least a version greater than v0.17. The Releases option in the UI is now disabled for CI/CD component projects.
With the GitLab agent for Kubernetes, you can share a single agent connection with a group. We aim to support a single agent across a large multi-tenant cluster. However, you might have faced a limitation on the number of connection sharing. Until now, an agent could be shared with only 100 projects and groups using CI/CD, and 100 projects and groups using the user_access keyword. In GitLab 17.0, the number of projects and groups you can share with is raised to 500.
If you need to run multiple agents in a cluster, we would like to hear your feedback in issue 454110.
In past releases, merge requests were tracked in a deployment only if the project’s merge method was Merge commit or Merge commit with semi-linear history. From GitLab 17.0, merge requests are tracked in deployments, including in projects with the merge method Fast-forward merge.
We published the following API Security Testing analyzer updates during the 17.0 release milestone:
- System environment variables are now passed from the CI runner to the custom Python scripts used for certain advanced scenarios (like request signing). This will make implementing these scenarios easier. See issue 457795 for more details.
- API Security containers now run as a non-root user, which improves flexibility and compliance. See issue 287702 for more details.
- Support for servers that only offer TLSv1.3 ciphers, which enables more customers to adopt API Security Testing. See issue 441470 for more details.
- Upgrade to Alpine 3.19, which addresses security vulnerabilities. See issue 456572 for more details.
Following the deprecation of Python 3.9 as the default Python image, Python 3.11 is now the default image.
As outlined in the deprecation notice, the target for the new default Python version was 3.10. The direct move to Python 3.11 was required to ensure FIPS compliance.
Secret Detection now uses an advanced vulnerability tracking algorithm to more accurately identify when the same secret has moved within a file due to refactoring or unrelated changes. A new finding is no longer created if:
A leak moves within a file.
A new leak of the same value appears within the same file.
Otherwise, the existing workflow (merge request widget, pipeline report, and vulnerability report) will treat the findings the same as before. By ensuring that duplicate vulnerabilities are not reported as secrets shift locations, teams are more easily able to manage leaked secrets.
GitLab Static Application Security Testing (SAST) now scans the same languages with fewer analyzers, offering a simpler, more customizable scan experience.
As announced, we’ve updated the SAST CI/CD template to reflect the new scanning coverage and to remove language-specific analyzer jobs that are no longer used.
You can now use the API to upload a custom avatar for any user type, including bot users. This can be especially helpful for visually distinguishing bot users, such as group and project access tokens or service accounts, from human users in the UI.
Thank you Phawin for your contribution!
As an instance administrator, when you use multiple browsers or different computers, it is difficult to know which sessions are in Admin Mode and which aren’t. Now, administrators can go to User Settings > Active Sessions to identify which sessions use Admin Mode.
Before this release, on self-managed GitLab, custom roles had to be created at the group level. This meant administrators could not centrally manage custom roles for the instance, which resulted in duplicate roles across the instance. Now custom roles are managed at the self-managed instance level. Only administrators can create custom roles, but both administrators and group Owners can assign these custom roles.
For more information on migrating existing custom roles, API endpoints, and workflows, see epic 11851.
This update does not impact custom role workflows on GitLab.com.
The security policy bot posts a comment on merge requests when they violate a policy to help users understand when policies are enforced on their project, when evaluation is completed, and if there are any violations blocking an MR, with guidance to resolve them. These comments are now optional and can be enabled or disabled within each policy. This gives organizations the flexibility and control to determine how they want to communicate about these policies to their users.
The GitLab Operator is now available for production use for cloud-native hybrid installations. See the installation documentation before adopting the GitLab Operator.
Support for a fallback to BusyBox images when you specify custom BusyBox values (global.busybox) is removed. Support for BusyBox-based init containers was deprecated in GitLab 16.2 (Helm chart 7.2) in favor of a common GitLab-based init image.
Support for gitlab.kas.privateApi.tls.enabled and gitlab.kas.privateApi.tls.secretName is also removed. You must use global.kas.tls.enabled and global.kas.tls.secretName instead.
The deprecated queue selector and negate options are removed from the Sidekiq chart.
Previously, members of groups that were invited to a group or project were visible only in the Groups tab of the Members page. This meant users had to check both the Groups and Members tabs to understand who has access to a certain group or project. Now, shared members are listed also in the Members tab, giving a complete overview of all the members that are part of a group or project at a glance.
Currently, most self-managed customers only utilize a single database.
In order to ensure that the setup between GitLab.com and self-managed is the same, we ask self-managed customers to migrate and run two databases by default.
In 16.0, two database connections became the default for self-managed installations.
In 17.0, we release two database mode as a limited Beta, with the goal to make running decomposed generally available by 19.0.
Migration to two databases remains optional in 17.0, but needs to be performed before upgrading to 19.0.
The migration requires downtime.
Self-managed customers can use a tool that executes this migration with some downtime.
We introduced a new gitlab-ctl command that allows you to upgrade your single-database GitLab instances to a decomposed setup.
This setup contains commands that will work with our Linux package.
The actual migration (copying the database) is part of a rake task in the GitLab project.
Bug fixes, performance improvements, and UI improvements
At GitLab, we’re dedicated to providing the best possible experience for our users. With every release, we work tirelessly to fix bugs, improve performance, and enhance UI. Whether you’re one of the over 1 million users on GitLab.com or using our platform elsewhere, we’re committed to making sure your time with us is smooth and seamless.
Click the links below to see all the bug fixes, performance enhancements, and UI improvements we’ve delivered in 17.0.
Our team members made significant contributions to Git 2.45.0. Here are a
few.
Reftables reference backend
Up to now, Git kept track of branches and tags (which both fall into the
category of “references” or “refs”) in loose files in the .git directory. Git can also
pack these up into a single “packed-refs” file. This can lead to:
Performance bottlenecks for those running monorepositories that have
many references.
Race conditions when references are updated or
deleted concurrently, leading to secondaries getting out of sync.
Reftables is a next generation
backend format for references that stores them not in loose files, but in
an indexed, binary file format. This format both scales with a large
number of references, and allows atomic updates.
The format was originally designed by Shawn Pearce in July 2017, and in
2021 a reftable library was upstreamed by Han-Wen Neinhuys, but was not
integrated into the Git project. The Gitaly team got a significant number
of changes merged including test refactoring, tooling, prepping the code
base to integrate the reftable backend, and many code optimizations to
the original implementation.
The result is that with Git 2.45.0, reftables is now usable! This
enables GitLab to run our Git servers with reftables as the references
backend for repositories, leading to improved performance for large
monorepositories.
Better tooling for references
With the introduction of the reftable backend, tooling was also needed
to inspect the references given it stores refs in a binary format.
Better tooling for references
Because the reftable backend stores references in a binary format rather than loose
files, we improved Git’s tooling for inspecting refs to allow users to
inspect refs when using reftable.
A new flag was added to git-for-each-ref(1) called --include-root-refs,
which will cause it to also list all references that exist in the root of
the reference-naming hierarchy. For example:
sh
$ git for-each-ref --include-root-refs
f32633d4d7da32ccc3827e90ecdc10570927c77d commit HEAD
f32633d4d7da32ccc3827e90ecdc10570927c77d commit MERGE_HEAD
f32633d4d7da32ccc3827e90ecdc10570927c77d commit refs/heads/main
Listing all reflogs
The reflog in Git tracks any
modifications to references and is useful for debugging what changed in a given
reference. Reflogs are kept in loose files in the .git/logs directory,
which can be inspected manually to list all log entries.
With the reftable format, these logs are no longer stored in the
.git/logs directory, making it impossible to inspect through the
filesystem.
Git lacked a command to list all reflogs in the directory. To address
this problem, a new command (git-reflog-list(1)) was added to list all reflogs,
which works with the reftable format.
More efficient packing of references
When the reftable backend is used to add or modify a ref, Git will
perform “auto-compaction”, which merges tables together as needed.
A --auto flag was added to git-pack-refs(1), which allows it to skip
reftable compaction when Git detects that the reference database is
already in an optimal state.
git-maintenance(1) has also been adapted to pass this flag when it calls
git-pack-refs(1) to make use of this new mode by default.
Before upgrading to GitLab 17.0, if you use Sentry to track errors in your GitLab environments, you must upgrade your Sentry instance to version 21.5.0 or later.
For more information, see the relevant documentation.
You must first upgrade to GitLab 16.11 before upgrading to GitLab 17.0 to allow background migrations to finish.
We want to hear from you
Enjoyed reading this blog post or have questions or feedback? Share your thoughts by creating a new topic in the GitLab community forum.
Share your feedback