[ci] Enabling aws-linux-scale-rocm to 10% #4899
Conversation
Add logic to configure_ci.py and configure_multi_arch_ci.py to select build runners using a weighted distribution: - Default builds: 90% azure-linux-scale-rocm, 10% aws-linux-scale-rocm - Sanitizer builds (asan/tsan): 100% azure-linux-scale-rocm-heavy-ramdisk Changes: - Add BUILD_RUNNER_LABELS config and select_build_runner() function to amdgpu_family_matrix.py - Add build_runs_on field to BuildConfig in configure_multi_arch_ci.py - Add build-runs-on field to matrix output in configure_ci.py - Pass build_runs_on through workflow chain to build stage jobs - Update multi_arch_build_portable_linux_artifacts.yml to use build_runs_on when provided, with fallback to existing logic Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add TestBuildRunnerSelection with 2 deterministic tests: - Test weighted selection (90% Azure, 10% AWS) for default builds - Test sanitizer builds always use Azure ramdisk runner Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
geomin12
requested review from
AmosLewis,
amd-justchen,
amd-shiraz and
jayhawk-commits
|
we hit the lottery :)
|
| sanity_check_only_for_family: ${{ matrix.family_info.sanity_check_only_for_family }} | ||
| release_type: ${{ inputs.release_type }} | ||
|
|
||
| build_python_packages: |
There was a problem hiding this comment.
what about this job here? also should get the dynamic cpu runners?
There was a problem hiding this comment.
we are doing an initial rollout to see any issues, but good, we'll do a 50% conversion and include other workflows
| sanity_check_only_for_family: ${{ matrix.family_info.sanity_check_only_for_family }} | ||
| release_type: ${{ inputs.release_type }} | ||
|
|
||
| build_python_packages: |
There was a problem hiding this comment.
same here for the future
There was a problem hiding this comment.
we haven't implemented windows yet for AWS runners, but will roll this out once we get that allocated
3 participants
As we are migrating from Azure -> AWS for build machines, we are increasing our CPU load from 10% for AWS, while 90% continues on Azure as we slowly migrate and test