Repository navigation
Recent stalling issues on MacOS 26 runner #13882
Description
Description
Since this week I experience stalling issues in my MacOS 26 x86_64 CI builds.
The workflow run seems to stall (hang) in some point of a compilation (C++) step (randomly is seems) and the workflow run gets terminated automatically after about an hour without any apparent errors.
Platforms affected
- Azure DevOps
- GitHub Actions - Standard Runners
- GitHub Actions - Larger Runners
Runner images affected
- Ubuntu 22.04
- Ubuntu 24.04
- Ubuntu Slim
- macOS 14
- macOS 14 Arm64
- macOS 15
- macOS 15 Arm64
- macOS 26
- macOS 26 Arm64
- Windows Server 2022
- Windows Server 2025
- Windows Server 2025 with Visual Studio 2026
Image version and build link
https://github.com/mcorino/wxRuby3/actions/runs/23909672830/job/69739799909
https://github.com/mcorino/wxRuby3/actions/runs/23922122359/job/69818505945
Is it regression?
Unknown. Did not ever see this problem before this week.
Expected behavior
The workflow run to finish normally, with or without errors.
Actual behavior
Workflow run seems to stall (hang) at random points in C++ compilation steps and gets terminated automatically with this notice:
"The hosted runner lost communication with the server. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error."
There is however no CPU or Memory starving action at that point in the workflow as far as I know.
Repro steps
Run the MacOS workflows.
Activity
hugovk commentedon Apr 3, 2026
We've been seeing similar things with
macos-26-intel:These might be the same or something else:
We've switched back to
macos-15-intelwhere we need to test on Intel.The Apple Silicon
macos-26seems okay.v-GeorgyPuzakov commentedon Apr 3, 2026
Hi @mcorino,
We will have a look and will back to you
SeanTAllen commentedon Apr 7, 2026
I'm seeing this constantly in the ponylang org.
v-GeorgyPuzakov commentedon May 13, 2026
Hi @mcorino,
We deployed new version for Mac 26, it might resolve this issue. Could you check this on your side?
mcorino commentedon May 13, 2026
AFAICT the problem has resolved. Thanks.
skirpichev commentedon Jun 9, 2026
Hmm, I still have this problem again:
https://github.com/diofant/python-gmp/actions/runs/27180789854/job/80240192639
Looks much less frequent, but it happens.
ci: add per-job timeout-minutes + ccache to dodge hosted-runner stalls
ci: add per-job timeout-minutes + ccache to dodge hosted-runner stalls
mcorino commentedon Jul 11, 2026
Ok, I encountered this issue on a MacOS 26 build again today. It seems this issue is definitely not solved yet.
ladvoc commentedon Jul 29, 2026
We are experiencing unexplainable test failures that only affect macOS 26 runners, particularly on timing sensitive tasks like receiving audio frames. The same tests pass on other platforms.
merge: break the mac CI cache ratchet and make a runner death legible
CI RED: the tolerated wedge CANCELLED the workflow instead of failing it
fix(ci): macos-amd64 从 macos-26-intel beta 镜像回退至稳定的 macos-15-intel
erayd commentedon Aug 27, 2026
Can confirm that I'm also currently encountering this issue. Random hangs midway through a test suite that passes 100% reliably when run on MacOS on my own machine, and passes 100% reliably using the
ubuntu-latestrunner on GitHub. Themacos-latestrunner fails more often than it passes, and always at a different point in my test suite. Re-running the failed job may occasionally pass, but usually just hangs again, at a different location in the test suite.Occasionally a stuck job may resume again on their own prior to timeout, but usually not. This does make me wonder if this is a resource starvation issue.
GloriousTacoo commentedon Sep 10, 2026
@v-GeorgyPuzakov Can this issue get reopened? It's pretty clear this problem has not been fixed.
feat(ci): mac x64 基底挪到 arm64 机器交叉编译，绕开 intel runner 停摆
feat(ci): mac x64 基底挪到 arm64 机器交叉编译，绕开 intel runner 停摆
fix(ci): avoid stalled Intel macOS 26 desktop runners (#2360)