Some builds or tests do require a significant amount of computing power and or RAM. A lot of time can be spared if more of both would be available.

So, starting now on https://travis-ci.com, it is.

vm:
  size: [large|x-large|2x-large]

• It is available for ‘full vm’ jobs only
• It is available for Linux, FreeBSD and Windows jobs on amd64 CPU architecture
• The arm64-graviton2 vm sizing is expected to arrive soon
• It is not available for OSX builds! The jobs on OSX will ignore the vm size tag
• It is available both on Concurrency and Usage based plans

Basic Travis CI build job utilizes virtual machine or LXD container with the computing power of 2 vCPU and 4 to 8 GB of RAM (depending on the scenario). Each subsequent vm size tier basically doubles the previous one e.g. large will spin up a virtual machine with 4vCPU and 16GB of RAM for your build job. Read more in our CI Environment Overview.

In order to use larger virtual machine instance sizes, you need to have credits in your account. The charge per started build minute is, as much as the resources, double of the previous tier.

Example: On a Concurrency Plan a Linux build job does not cost credits, as this is already covered by a monthly subscription. On a Usage based plan the basic cost of a Linux build minute is 10 credits. Using a large virtual machine instance (doubling the resource) will require 20 credits for each started minute in both scenarios.

Read more on cost in our Billing Overview.

Free Plan has the instance sizes available, so you can take it for a test ride and see how it fits your scenario.

Why not variable size for LXD containers?

It’s already there, assigned automatically if resources are available. LXD containers are assigned a minimum resource pool which can be dynamically increased by the LXD host during their lifetime, depending on how noisy the current build job neighborhood is. At the moment LXD containers at Travis CI always start with a computing time equal to 2 vCPU and a certain amount of RAM. Whenever there’s a spare resource in LXD host, the container may receive it. Therefore our LXD container build jobs (namely for different CPU architectures) are ignoring the vm instance size tag.

What about the AWS Arm64 Graviton2 ‘full vm’?

The variable instance size for your Arm64 Graviton2 build jobs is in works and should be available for you in a reasonable timeframe. Stay tuned to our announcements and blog post.

We hope the availability of larger vm instances will make your builds faster, more convenient and more effective.

Happy building!