Nov 05, 2026
Resolved: arm64 Macro Runners Outage
From midnight UTC on November 5, 2026 until Nov 5, 20:00 UTC, our arm64 macro runners were unavailable. Our infrastructure provider removed our access to the instances they run on by mistake, so benchmark jobs targeting them failed or stayed queued. Access is restored and arm64 jobs run normally again.
Updates
- Nov 5, 00:00 UTC — arm64 macro runner jobs stop starting.
- Nov 5, 10:30 UTC — Outage confirmed. We're contacting impacted teams directly and working on a solution.
- Nov 5, 19:30 UTC — Our infrastructure provider confirmed the removal was an error on their side and has started restoring access. arm64 jobs will resume automatically once it’s effective.
- Nov 5, 20:00 UTC — Resolved. Access to the arm64 instances is restored and jobs on the arm64 labels run normally again.
Who was affected
Workflows using one of these runner labels:
runs-on: codspeed-macro-arm64-graviton-ubuntu-22-04
# or the former label
runs-on: codspeed-macro
Not affected:
- The x86 macro runners on GitHub Actions
(
codspeed-macro-x64-ryzen-9950x-ubuntu-24-04) - CircleCI macro runners, which already run on the x86 machines
- CPU simulation and walltime benchmarks on standard CI runners
Failed jobs during the outage were not billed.
What you need to do
Nothing. Jobs on codspeed-macro-arm64-graviton-ubuntu-22-04 (and codspeed-macro)
run again with no change to your workflows, and your existing baselines are still
valid. If a job queued during the outage never ran, re-run it.
We still plan to move everyone to our x86 macro runners. We'll share the timeline and the new label separately, so you can switch when it suits you.
Questions in the meantime: reach out on our Discord or at [email protected].
We're sorry for the disruption, and thanks for your patience.
We'll publish a postmortem with the full details of what happened and why.








