Spawn Hosts - turalf/evergreen GitHub Wiki
If a test fails on a platform other than the one you develop on locally, you'll likely want to get access to a machine of that type in order to investigate the source of the failure. You can accomplish this using the spawn hosts feature of evergreen.
Evergreen administrators can choose to make a distro available to users for spawning by checking the box on the distro configuration panel labeled "Allow users to spawn these hosts for personal use"
Only distros backed by a provider that supports dynamically spinning up new hosts (static hosts, of course, do not) allow this option.
Visit /spawn to view the spawn hosts control panel. Click on "Spawn Host" and choose the distro you want to spawn, and choose the key you'd like to use (or provide a new one).
Alternately, for a task that ran on a distro where spawning is enabled, you will see a "Spawn..." or "Spawn Host" link on its task page.

Clicking it will pre-populate the spawn host page with a request to spawn a host of that distro, along with the option to fetch binaries and artifacts associated with the task and any tasks that it depended on.

Fetching artifacts can also be performed manually; see fetch in the Evergreen command line tool documentation.
Artifacts are placed in /data/mci. Note that you will likely be able to ssh into the host before the artifacts are finished fetching.
If your project has a project setup script defined at the admin level, you can also check "Use project-specific setup script defined at ..." before creating the spawn host. You can check if there are errors fetching artifacts or running this script on the host page: https://spruce.mongodb.com/host/<host_id>.
EC2 spawn hosts can be stopped/started and modified from the Spawn Host page, or via the command line, which is documented in Basic Host Usage in the Evergreen command line tool documentation.