You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
The amd64 Drone runner used for PRs is a AMD EPYC 7402P, 24/48 Core/Thread processor. When we run the amd64 build pipeline, we utilize less than 25% of the CPU and only 10GB of the 64GB of RAM. The Drone pipelines are currently too sequential to full utilize this resource.
Describe the solution you'd like
Increase parallelization of Drone build steps where possible. Consider:
Allowing multiple k3s build pipelines at once
Run more E2E tests simultaneously
Run more Docker test simultaneously
Describe alternatives you've considered
We have moved most of our workloads to GHA. But these machines are many, small 4vCPU/16GB nodes. Drone runners enable large scale testing and performance, and should be utilized as such.
Additional context
The text was updated successfully, but these errors were encountered:
In the past, increasing test concurrency caused problems with disk IO latency - builds running in parallel with datastore ops and image pulls on the same underlying disk caused problems and we saw a lot of tests running long and flaking. I don't think CPU or memory were the bottleneck.
Is your feature request related to a problem? Please describe.
The amd64 Drone runner used for PRs is a AMD EPYC 7402P, 24/48 Core/Thread processor. When we run the amd64 build pipeline, we utilize less than 25% of the CPU and only 10GB of the 64GB of RAM. The Drone pipelines are currently too sequential to full utilize this resource.
Describe the solution you'd like
Increase parallelization of Drone build steps where possible. Consider:
Describe alternatives you've considered
We have moved most of our workloads to GHA. But these machines are many, small 4vCPU/16GB nodes. Drone runners enable large scale testing and performance, and should be utilized as such.
Additional context
The text was updated successfully, but these errors were encountered: