This update fixes ingestion of large tuist inspect test reports and hardens the xcresult processor VMs.
Test results ingestion
- Large xcresult bundles with thousands of test cases now upload reliably. Previously, the server looked up existing test cases with one parameter per ID, which hit ClickHouse limits as a “Too many form fields” error. The lookup now passes all IDs as a single
Array(UUID)parameter, so the form-field count stays constant regardless of report size. The batch size is also lowered to keep the array value under ClickHouse’s per-field length limit.
xcresult processor VMs
- Stopped a disk leak caused by attachment temporary directories left behind after each processing run. Attachments are now exported under the worker’s per-run scratch directory, which is already cleaned up at the end of the job.
- Restored VM metrics forwarding so the dashboard and Prometheus can see worker activity again after the Tailscale migration.
- Added guest disk monitoring. A new
DiskPressurenode condition reports when a guest VM’s root volume reaches 90% or more, and atart_kubelet_guest_disk_usage_percentgauge is exported so you can alert on disk pressure before jobs run out of space.