You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
iOS tests failing with OSError: [Errno 28] No space left on device #6473
Since this known issue is recurring across many WorkItemExecution legs, I would treat it as a per-work-item disk budget problem rather than only matching the final OSError: [Errno 28] line.
The useful split is to capture sizes before the iOS work item starts and again on failure, then separate:
df -h
du -xhd1 "$HELIX_WORKITEM_ROOT""$TMPDIR""$HOME/Library/Developer"2>/dev/null | sort -h | tail -60
find "$HELIX_WORKITEM_ROOT" -xdev -type f -size +250M -print 2>/dev/null | head -80
For iOS/macOS legs I would also bucket DerivedData, simulator data, NuGet/package caches, test payloads, crash dumps, and any copied app bundles separately. Those have different cleanup risk; deleting simulator runtimes or shared tool state is much riskier than rotating per-work-item temp/output.
I made a read-only macOS CI storage scan/checklist for exactly this style of runner evidence collection. It does not delete anything; it just produces the before/after artifact needed to decide safe vs review-first cleanup:
The important signal for this issue is whether the repeated hits are dominated by one growing work item payload, shared Xcode/simulator state, or temp output that is not being removed between attempts.
Activity
liuminsheng3 commentedon Jun 16, 2026
Since this known issue is recurring across many
WorkItemExecutionlegs, I would treat it as a per-work-item disk budget problem rather than only matching the finalOSError: [Errno 28]line.The useful split is to capture sizes before the iOS work item starts and again on failure, then separate:
For iOS/macOS legs I would also bucket DerivedData, simulator data, NuGet/package caches, test payloads, crash dumps, and any copied app bundles separately. Those have different cleanup risk; deleting simulator runtimes or shared tool state is much riskier than rotating per-work-item temp/output.
I made a read-only macOS CI storage scan/checklist for exactly this style of runner evidence collection. It does not delete anything; it just produces the before/after artifact needed to decide safe vs review-first cleanup:
https://site-iota-ebon-80.vercel.app/ci-macos-storage-scan.html?utm_source=github&utm_medium=issue-comment&utm_campaign=ci-macos-storage&utm_term=dotnet-ios-tests-errno28&utm_content=dotnet-dnceng-6473
The important signal for this issue is whether the repeated hits are dominated by one growing work item payload, shared Xcode/simulator state, or temp output that is not being removed between attempts.