Skip to content

Solutions

Labels and pools: Standard vs High-Power

Your workflow's runs-on labels choose the pool: Standard for everyday machines, High-Power for larger, faster ones. Each pool has its own balance.

Symptom

You are setting up your workflows and want to choose between Standard and High-Power runners — or your jobs are not reaching SolveLang at all and simply wait in the queue.

Cause

GitHub routes every job by the labels in its runs-on line. SolveLang runners answer to specific label sets. A job that asks for labels SolveLang does not answer to will wait for a different runner — that is not an error; the job is simply not addressed to SolveLang.

Fix

Use one of these label sets in your workflow's runs-on:

  1. For Standard runners — everyday machines for ordinary jobs: ["self-hosted", "solvelang-runners-standard"].
  2. For High-Power runners — larger, faster machines for heavy jobs: ["self-hosted", "solvelang-runners-high-power"].
  3. Plain ["self-hosted"] also reaches SolveLang, and runs on the Standard pool.
  4. Other labels your workflow already carries — for example, labels that pick out your own runners — can ride alongside a SolveLang label in the same runs-on line, and the job still runs on the pool the SolveLang label names; a runs-on line with no SolveLang label at all is not picked up by SolveLang.
  5. Remember that the pools are separate balances: the pool a job runs on is the balance it draws from. A Standard job never spends High-Power minutes, and a High-Power job never spends Standard minutes.
  6. Speed is never free, and there are two ways to buy it: High-Power minutes put a job on larger, faster premium machines at the High-Power prices, and On-demand capacity starts a job right away while consuming minutes at 2×. Standard jobs on Spot capacity bill at the plain 1× rate.

Example

The same workflow can mix pools, job by job:

test:
  runs-on: ["self-hosted", "solvelang-runners-standard"]

heavy-test:
  runs-on: ["self-hosted", "solvelang-runners-high-power"]