See the issue reported https://gitlab.com/redhat/centos-stream/ci-cd/zuul/jobs-config/-/issues/1

Some package build root requires more diskspace than available on the test node. We use the t2.medium flavor on EC2 which is by default 5GB root FS.

Solution could be:

• always attach an EBS volume to the node. The Nodepool AWS driver provide volume-type and volume-size (https://zuul-ci.org/docs/nodepool/aws.html).
• Add a new Nodepool label such as cloud-fedora-34-aws-edisk. Distgits that needs more diskspace than default can opt-in for this image.

It seems we can either ask to attach an EBS volume (through nodepool label options) or use some instance flavor that already provide a SSD such as m5d.large.