Users can now orchestrate their model deployments across multiple environments through the Amazon SageMaker Orb. They can also track and manage their deployments from within the Releases UI.

See our documentation for more information.

Collapse

Inbound Webhooks are now available in open preview.

With this new feature, you can trigger a pipeline from anywhere that emits a webhook or run a curl command. With this change, we also updated pages such as the Dashboard to accurately reflect that pipelines can now come from outside your VCS.

Only available to users authenticated through our GitHub App integration. See our community forum and an AI use case for more.

Collapse

CircleCI’s “Linux CUDA” image has been updated to include the Pytorch, Tensorflow, and HuggingFace transformers libraries out-of-the-box.

Users will no longer need to waste time pulling these large libraries into their execution environment when a job is running. This image is currently only available to jobs that use the gen2 GPU resource class.

To use this new image, specify:

linux-cuda-12:default or linux-cuda-12:edge in the image stanza.

Collapse

Now in preview, three new resource classes that use GPUs:

1. gpu.nvidia.small.gen2: uses 1 NVIDIA Tesla A10G GPU (4 CPU / 16 GB RAM 200 credits/min)
2. gpu.nvidia.small.multi: uses 2 NVIDIA Tesla T4 GPUs (4 CPU / 16 GB RAM 160 credits/min)
3. gpu.nvidia.medium.multi: uses 4 NVIDIA Tesla T4 GPUs (8 CPU / 30 GB RAM / 240 credits/min)

Only available to users on our Scale pricing plan. See our community forum for more.

Collapse

Machine runner 3.0 is now Generally Available, providing improvements to usability, setup, and maintenance of machine runners.

Container runners and machine runners are now built using the same framework. Launch agent users will need to upgrade to the new application, but existing container runner installations will not require any additional action.

Faster feature enhancements: We’ll now bring many network resilience feature enhancements to both types of self-hosted runners (machine and container).

Improved usability: With more uniform logs and metrics, CircleCI’s machine runners ensure consistent and reliable performance and a seamless experience for users regardless of their runner setup.

Setup and maintenance: CircleCI’s machine runner 3.0 simplifies the setup and maintenance process by offering convenient packages that allow you to easily integrate with your preferred cloud templating tools.

Linux, macOS, and Windows support is available. See all the details on the runner documentation.

Machine Runner 1.1 variants will be supported through March of 2024, giving you time to plan a migration.

View our documentation to upgrade.

Collapse

CircleCI is upgrading the Docker version that is used for jobs that use the Docker executor in November, 2023. This change is critical to ensuring the underlying infrastructure that runs your jobs continues to be secure, reliable, and performant.

This upgrade should be effectively invisible to customers, no action is needed. The upgrade is from Docker 20.10.23 to Docker 24.

See our community forum for more details.

Collapse

Before Upgrading

See the CircleCI server 4.3 release notes and upgrade guide for this release.

There is a background database migration job that is run on install/upgrade to server 4.3. Customers upgrading to 4.3 must ensure this job completes. Full details can be found in the Upgrade guide.

When upgrading from server v3.x you will need to install v4.0 and then follow the recommended upgrade path before installing this version.

What’s New in Release 4.3.0

The v4.3 release introduces upgraded execution architecture. Previously, VM service was responsible for creating virtual machines for machine and remote Docker jobs. This has been replaced with machine provisioner. The new architecture has improved efficiency and stability. This upgrade also brings server inline with CircleCI cloud execution, meaning server will benefit from future updates faster. Some specific benefits include:

• A new UI for step output, which is better at handling larger output and colour.
• New systems for processing job output (step data, caches, artifacts etc.) is much more efficient.
• Machine and remote Docker jobs no longer run through Nomad so compute costs are cut down significantly.
• You do not need to run a Nomad cluster if you just want to run machine (and/or remote Docker jobs) on your server instance.
• DLC is now configured to run through S3, which is cheaper than running on SSD volumes, and there is no longer a limit on the amount of DLC jobs that can be run in parallel. DLC stays more consistent with better pruning and is shared between all jobs in a project instead of each cache being different.
• VM service could fall back to another region/zone, but machine provisioner can handle gaps in availability through:
  • Distributing load across availability zones through round-robining
  • Sending a percentage of traffic to a different zone
  • Instance type fallbacks based on deficits and success

NEW FEATURES

• Server execution architecture has been upgraded from VM service to machine provisioner. For further information, see the Installation guide.
• Server now ships with a custom postgreSQL image which now supports pg_crypto, pg_partman, and pg_cron support extensions.
• Added a REPL toggle to turn on a background which deletes workflow data older than 1 year. This is off by default.
• Config policies are now fully supported on server. _compiled_ is now a reserved word in config policies, so that parameterized configuration is fully available to the config policies engine. For further information see the Config policies overview docs.
• The circleci env subst command is now supported. See the Environment variables docs for more information.
• Added support overriding non-CircleCI-managed images.

CHANGES

• Nomad Autoscaler has been updated to 0.3.7.
• Server no longer deploys the Prometheus chart.
• A new service, Runner admin, has been added to server.

FIXES

• Fixed a bug that caused the Insights page to crash if a user tried to view job resource usage, which is unsupported on server at this time.
• IF a job is retried due to an infrastructure failure, it should no longer cause a workflow to get stuck

KNOWN ISSUES

• Vault may not refresh its client token after a month of uptime.
• Retry with SSH for jobs using the machine executor advertises a private IP address. For this reason, retry with SSH for jobs using the machine executor works as standard for public installations, but for private installs you would need to ensure that you can access the private IP advertised. For example, by using a VPN into your VPC.
• CircleCI 1.0 builds are not supported. If an attempt is made to run a 1.0 build, no feedback will be available in the application to indicate the cause of the issue. If a build is run on your installation and does not show up in the CircleCI application, users should be directed to use the CircleCI CLI to validate the project configuration and get details of the possible cause of the issue.

To learn more about Server 4.x installation, migration, or operations please see our documentation.

Collapse