The latest release of the Buildkite Terraform provider adds buildkite_meta: a data source providing the IP addresses Buildkite uses for webhooks ✨

The new data source can be used in your terraform project to dynamically set firewall and ingress rules, allowing in traffic from Buildkite 🚦

You can find a code sample and attribute reference in the Buildkite provider docs on the Terraform Registry. For more detailed information about the addition, check out the v0.4.0 release on GitHub.

To support customers with inbound network firewalls, Buildkite's Meta API lists the IP addresses that Buildkite uses to send outbound traffic such as webhooks and commit statuses.

To improve network resilience, new IP addresses were added to the Meta API earlier this month. If you use network ACL/firewall rules to allow-list traffic from Buildkite, these should be updated immediately.

During the week of 2021-04-19 traffic will intermittently be switched to the new IPs for short periods as a low-impact method of uncovering issues. During the following week of 2021-04-26 outbound traffic will be permanently switched to the new addresses.

If you only want to build tags, you can now optionally switch off building branches 🚦

We've added a new setting: Build branches that is on by default for new pipelines. To only build on tags, disable Build branches and ensure that your Build tags box is checked:

You can find the Build tags and the new Build branches setting in your pipeline's Repository settings when using GitHub, GitHub Enterprise, Bitbucket, or Bitbucket Server.

We've heard that some Buildkite customers are considering ARM in production, so we've made sure that Elastic Stack for AWS v5.1.0 can help you get there ✨

v5.1.0 adds support for the following AWS instance types with an ARM-powered Graviton CPU:

• m6g.*
• m6gd.*
• t4g.*
• a1.*
• c6g.*
• c6gd.*
• r6g.*
• t6gd.*

To use these instance types, follow the upgrade guide for v5.1.0 and set your InstanceType stack parameter 🚀

Apple machines with the new ARM-based M1 chip are being delivered to Buildkite customers, and starting from agent 3.26.0 we have an experimental binary that will run natively on these systems.

Users with M1 based hardware can install the native binary by:

• Using Homebrew
• Running our curl install script
• Downloading the darwin-arm64 binary directly from GitHub

As part of the transition to GraphQL, we have added graphql_id for the following REST API resources. The graphql_id is used to target specific resources when running queries and mutations on GraphQL API.

• Agent
• Build
• Job
• Organization
• Pipeline
• Team
• User

Here is an example response from pipelines which includes graphql_id:

1
2
3
4
5
{
  "id": "849411f9-9e6d-4739-a0d8-e247088e9b52",
  "graphql_id": "UGlwZWxpbmUtLS1lOTM4ZGQxYy03MDgwLTQ4ZmQtOGQyMC0yNmQ4M2E0ZjNkNDg=",
  "url": "https://api.buildkite.com/v2/organizations/acme-inc/pipelines/my-pipeline"
}

If you have any questions or feedback, please let us know in the Forum or send us an email 📨

With the new block step attribute, blocked_state, you can specify what state the build should have when it’s blocked by each block step — green passed, yellow running, or red failed.

When you’ve got a block step that requires user input, having the build look or behave like it's passed isn’t always what you want. Setting blocked_state will affect the build state within the Buildkite UI, source provider integrations, webhooks, and notifications. This gives you more control over how your builds interact with PRs and custom integrations.

Add the new blocked_state attribute to your block steps to change the blocked status colour from the default green.

You can now mix positive and negative branch filter patterns for more precise control over which branches to build 🔀

We've done our best to reach out via email if your pipeline already had a mix of patterns that will be affected by this change 👋🏻

If you're seeing unexpected builds, or are left waiting for a build that isn't showing up, we'd suggest a quick review of your pipeline's branch filter patterns.

For more information on branch filter patterns and some example code, check out the updated branch configuration docs.

We've released a new major version of the Elastic CI Stack 🌈

The latest version includes:

• Experimental Windows Server 2019 support
• Support for four new regions: ap-east-1 (Hong Kong), me-south-1 (Bahrain), af-south-1 (Cape Town), eu-south-1 (Milan)
• Support for managing Agent tokens with AWS SSM
• A blazing fast Agent Auto Scaler by AWS Lambda - up to 300% faster initial scale-ups compares to previous AutoScaling rules

And much, much more! Check out the full changelog for this release on GitHub.

We'd love for you to share any feedback you have on the v5 announcement post in our forum 🤓

Thank you to the many users who provided bug reports and contributions that made this release possible 💚

Debugging job and agent issues has become a whole lot easier with the shiny updates we've just released to the Timeline tab 📖

The new and improved Timeline tab has:

• Live updating with agent events
• Events from the dispatcher showing when the jobs are scheduled and assigned, and
• All your job info in one place including concurrency groups, step dependencies, and conditionals

You can find all of these updates under the Timeline tab of any job 🕵🏻‍♀️

See exactly which repo permissions are requested by Buildkite with our new GitHub App

GitHub Apps create a direct connection between your Buildkite and GitHub organizations, removing reliance on individual user connections. GitHub Organization Admins can then see permissions and manage access on a per-repo basis 🙌🏻

Get started by installing the GitHub App from your Repository Providers page.

We've made it easier to manage multiple source control integrations 🔀 You can now find a list of connected and available integrations on the new Repository Providers page

Click through to each provider for settings and delete options ⚙️

Skipped builds are now marked as an "error" in GitHub commit statuses:

Build skipping, available from your Buildkite pipeline settings, saves you time by only testing the latest commit on a branch. Previously, a skipped build created a "success" commit status on GitHub, but this could give the false impression that the tests were run and passed.

The options here are pretty limited, but we've made this change to better reflect that the build is no longer pending, but didn't succeed or fail, it finished unusually.

Running another build on the same commit will replace this error status, and running builds on future commits will correctly update a pull request as expected.

We’ve taken some steps to ensure that the Buildkite Agent is ready for upcoming Macs running on Apple Silicon. 👩🏻‍💻✨

First things first, we are eagerly anticipating Go, the language the Agent is written in, gaining the ability to build macOS binaries specifically for Apple Silicon, and plan to provide such binaries as soon as it’s feasible to do so.

Currently we suggest the 64-bit Intel amd64 binary, which happily runs under Rosetta 2 on Apple Silicon. 🦾

For those already adventuring on early Apple Silicon Macs, we’ve verified that our Homebrew-based Mac installation instructions work just fine there. Our non-Homebrew and Linux install script has also been updated to ensure it picks the correct binary on Apple Silicon Macs.

Both of these installation options will install the 64-bit Intel binary, and while the Agent itself will run within Rosetta 2, running an Apple Silicon binary or test suite within your builds should work just fine. We don’t anticipate any obvious issues with this configuration, but please report any issues you do encounter to us.

We have also decided to stop providing 32-bit Intel 386 binaries for macOS as of the next Agent release. This is intended to avoid confusion due to the fact that neither current Macs running macOS Catalina nor upcoming Apple Silicon Macs can run them. For those curious, since early 2018, only the 64-bit Intel binaries have been supplied via our Homebrew formula. If you have a specific need for a 32-bit Intel Mac binary, building the Agent yourself will likely work for the foreseeable future.

To the Linux, BSD and Windows users out there; we have no plans to stop providing 32-bit Intel binaries for your platforms. ✌🏼

If you have any questions we haven’t covered here, or run into some issues in your continuing adventures, please let us know via support@buildkite.com.

We've added new documentation on how to run Bazel on Buildkite, including a GitHub repository and example pipeline 📗

Check out the new Using Bazel on Buildkite documentation for details, with links to the example repository and further reading.

Since adding conditionals support we've added the following 11 new variables and one new function 🎉

• build.env()
• build.author.email
• build.author.id
• build.author.name
• build.author.teams
• build.commit
• build.creator.email
• build.creator.id
• build.creator.name
• build.creator.teams
• build.pull_request.draft
• build.source

You can find full descriptions of all the available variables in the Using Conditionals guide, as well as new code samples 💫

Buildkite can send notifications to your favourite Slack channel when your builds have finished. These notifications can be filtered in a few different ways, including to only "failed" and "fixed" builds. Historically, "fixed" meant there was a previous build on the same branch which failed and the new build has passed. Now "fixed" is also sent when a new build becomes blocked.

We've tried to make this change carefully so that existing notifications are mostly unaffected, apart from "fixed" notifications appearing when builds become blocked. If this interrupts your workflow, please reach out via support@buildkite.com.

Previously, the build would consider a timed out job as "failed", regardless of its exit status 🙃

This could sometimes happen if a command finished just as the job was being timed out, or if it had specifically implemented signal handling to gracefully exit.

Now, jobs that timeout with an exit status of 0 will be marked as "passed", so you can depend on the exit status as the one source of truth for job status ✅

You can now more easily embed colored terminal output in annotations. 🌈

Wrap any ANSI formatted console output in a Markdown block with either the term or terminal syntax:

1
2
3
```term
Fancy \x1b[91mc\x1b[33mo\x1b[93ml\x1b[92mo\x1b[94mr\x1b[95ms\x1b[0m here
```

The ANSI formatting is then rendered for you in the annotation:

You can read more about the formatting supported in annotations in the CLI docs.

Builds created from schedules will no longer be associated with users 🏃🏼‍♂️💨

If a user loses access to a pipeline, either by leaving the organization or changing teams, you won't need to update your schedules to a new owner.

If your schedule is still attached to a user, you'll see the ⚠️ Requires Migration badge. You'll be able to remove the user from the schedule on the build schedule page.

Unmigrated schedules will have their users automatically removed on the 30th of January, 2020. To check your build schedules, see the Schedules section of your Pipeline Settings 👀