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
The 4.22-virt-kdm-e2e-test-aws job failed while checking payload B after a successful incremental CBT backup. ChecksumBlockDeviceRegion received empty stdout from the read-only dd ... | sha256sum pod exec and returned sha256sum produced no output for /dev/volume0 region in alpine-guestagent/virt-launcher-alpine-guestagent-z5796. The same region was checksummed successfully before the backup, and payload A was checksummed successfully immediately before this failure.
This tracks an intermittent E2E checksum-exec failure, not a confirmed backup data-integrity defect or a confirmed API-server stream defect. The reported PR changes only BSL region handling.
Reproduction evidence
Job: pull-ci-openshift-oadp-operator-oadp-1.6-4.22-virt-kdm-e2e-test-aws; build ID 2102783124353060864; OCP 4.22.0-0.nightly-2026-09-22-003645.
Failed test: restore from an incremental (not full) kubevirt-datamover CBT backup in tests/e2e/virt_backup_restore_suite_test.go (Line 1841).
tests/e2e/lib/virt_helpers.go: ChecksumBlockDeviceRegion rejects empty stdout; execShellCommandInPodWithRetry returns immediately when exec returns no error. Its current retry predicate checks for dial tcp errors only.
Check the CI artifacts and reruns for recurrence. Capture stdout, stderr, exec errors, and pod status to distinguish a missing exec stream from a command failure.
If an empty successful checksum exec is transient, add a bounded retry for this read-only checksum and a focused test. Keep real command failures visible.
Verify that valid checksums still pass and that persistent empty output still fails with useful diagnostics.
The separately reported kubevirt/kubevirt#18957 SourceLost race is not the checksum failure tracked here. Related E2E exec work: #2451.
Activity