Evolving Bitbucket Pipelines to unlock faster performance and larger builds

Bitbucket Pipelines has seen amazing adoption over recent years, with millions of developers and teams using it to build better software every day.

As part of our ongoing commitment to improving performance and reliability, we are in the process of making enhancements to our cloud infrastructure.

These changes will also unlock a range of powerful new capabilities for customers that will turbocharge their CI/CD workloads and empower engineers to focus more of their energy on building great software.

Faster builds and more memory — powered by a new container runtime

As part of these changes, we are moving to a new runtime under the hood. This new runtime uses microVM’s and runc to give us better capabilities in regard to how we assign compute resources to steps. Thanks to these changes, we are excited to announce the upcoming availability of 4x and 8x sized steps, releasing in June.

In addition to these larger step sizes, we are making changes to how compute resources scale with larger sized steps in order to unlock faster build speeds for our customers. In the past, when larger step sizes were used (2x vs 1x), the memory available to the build increased, but the amount of CPU remained the same. However, thanks to our new cloud runtime, we are now also able to scale the number of CPU cores available to larger step sizes (4x & 8x), in addition to the memory, more details will be shared in the coming weeks.

This runtime will initially only be available to 4x and 8x steps, with 1x and 2x steps being migrated over during August.

Changes to Pipelines IP address ranges

To enable the adoption of the new runtime, we are making changes to how IP addresses are assigned to the nodes that steps execute on.

If your team or organization uses a firewall, access list, security group or other network policy to restrict incoming access to particular IP's or IP ranges, you may need to update those rules to permit connections from our new addresses.

How will this affect you

NOTE: There will not be any downtime for this migration, and most users will not need to make any changes. However, there are some changes that may be required in highly specialised setups.

For existing 1x and 2x steps

There are no immediate changes for 1x and 2x steps. However, these step sizes will be migrated to the new runtime during August. At which point they will transition to run on nodes restricted to IP addresses provided by our cloud partner AWS.

These addresses can be found by using the following endpoint and using the us-east-1 and us-west-2 regions.

In order to avoid any disruptions in your Pipelines usage once the migration occurs during August, any allow-list policies in use should be updated to include these IP addresses.

For new 4x and 8x step users

As the new 4x and 8x steps use the new runtime by default, they will run on nodes that use IP addresses provided by our cloud partner AWS.

These addresses can be found by using the following endpoint and using the us-east-1 and us-west-2 regions.

Any allow-list policies in use should be updated to include these IP addresses in order to take advantage of the new step sizes.

What if I need a more restrictive range of IP addresses?

We understand that in some highly sensitive scenarios, a more restrictive set of IP addresses may be desired in order to limit the range of IP's that are allow-listed in your firewall.

To support this, we are adding an option for 4x & 8x steps that will allow you to execute those steps on a dedicated pool of nodes, restricted to a narrower sub-set of persistent IP addresses.

Note: This option is only available to 4x & 8x steps