Runner pool
Pick a runner with tags, not by name — names change when hosts are recycled.
Linux, x86-64
| Runner | Tags | vCPU / RAM | Load | State |
|---|---|---|---|---|
| ams-l-03 | linux, docker, node | 8 / 16 GB | 72% | busy |
| ams-l-05 | linux, docker, go | 8 / 16 GB | 11% | idle |
| ams-l-07 | linux, docker, node, python | 16 / 32 GB | 64% | busy |
| ams-l-09 | linux, docker, node | 8 / 16 GB | 38% | busy |
| ams-l-11 | linux, docker, go, rust | 16 / 32 GB | 7% | idle |
| ams-l-14 | linux, docker | 4 / 8 GB | 0% | drained |
Image builders
| Runner | Tags | Storage | Queue | State |
|---|---|---|---|---|
| ams-x-01 | image, buildkit, scan | 1.2 TB NVMe | 2 | busy |
| ams-x-02 | image, buildkit, scan | 1.2 TB NVMe | 0 | idle |
ams-l-14 is drained for a kernel upgrade. Jobs pinned to that name will
queue indefinitely — use tags.
Capacity notes
- Peak load is 09:00–11:00 CET, right after the nightly merge window.
- E2E suites are capped at 4 concurrent jobs to keep the browser pool responsive.
- Image builds run on NVMe hosts only — buildkit cache is not shared with Linux runners.