| name | openvmm-ci-investigation |
| description | Investigate CI failures on OpenVMM PRs. Load when a PR has failing CI checks, you need to download and analyze test artifacts, or you need to diagnose build, fmt, clippy, or VMM test failures. |
Investigating CI Failures
When a PR has failing CI checks, always start by running the investigation
script. Do not manually query the GitHub API or download artifacts by hand
— the script handles all of that automatically.
Step 1: Run the Script
python3 repo_support/investigate_ci.py
python3 repo_support/investigate_ci.py 2946
python3 repo_support/investigate_ci.py 23017249697
The script automatically:
- Resolves the PR to every failed CI run for its head commit. A commit
can have multiple workflows (e.g. "OpenVMM PR" and "OpenVMM Docs PR"),
and each failing one is investigated in turn — so don't assume a single
run covers everything. Only the newest run of each workflow is used, so
superseded runs are ignored; a
NOTE: <workflow> is still in_progress
line means results are incomplete and more failures may appear later.
- Identifies failed jobs
- Downloads
*-unit-tests-junit-xml artifacts and parses JUnit XML for
unit test failures
- Downloads
*-vmm-tests-logs artifacts if they exist
- Finds tests with
petri.failed markers and extracts ERROR/WARN lines
- If no test or JUnit artifacts exist (build/fmt/clippy failure), shows the
tail of the failed job's log
Step 2: Analyze the Script Output
Read the script's output to identify:
- Which tests failed (unit tests and/or VMM tests)
- Error messages and root causes
- Whether it's a build/fmt/clippy failure vs. a test failure
When more than one run failed, the output is split by # RUN n of m banners.
Check all of them — a docs failure and a build failure are reported in
separate runs.
Then use the information to diagnose the issue and suggest fixes.
Step 3: Diagnose from Logs and Code
Always try to diagnose the failure from CI logs, test code, and error
messages first. Most failures can be understood without local reproduction.
Read the relevant test source, trace the error through the code, and form
a hypothesis before considering local repro.
Step 4: Reproduce Locally (Only If Needed)
If you cannot diagnose the failure from logs alone, ask the user whether
they want to attempt local reproduction before proceeding. Do not
automatically start building or running tests on the user's machine.
If the user agrees, check whether the failing platform matches the host
architecture — you can only reproduce tests locally on the same arch. If
it doesn't match, explain this to the user and continue diagnosing from
CI logs and test code.
To reproduce locally, load the vmm-tests skill for instructions on
running with cargo xflowey vmm-tests-run. Do not use
cargo nextest run -p vmm_tests directly — it won't have the required
artifacts.
Reference: Manual Commands
Only use these if the script fails or you need to dig deeper into a
specific artifact. In normal usage, the script above is sufficient.
If the script isn't available or you need more control, follow these steps:
1. Find the failing run
gh pr checks <PR_NUMBER> -R microsoft/openvmm
gh run list -R microsoft/openvmm --commit <SHA>
2. Identify the failing job
gh run view <RUN_ID> -R microsoft/openvmm --json jobs \
-q '[.jobs[] | select(.conclusion == "failure") | {name, databaseId}]'
3. Download test artifacts
Unit test results are stored in JUnit XML artifacts named
{platform}-unit-tests-junit-xml. Known platforms include:
x64-linux
aarch64-linux
aarch64-linux-musl
gh run download <RUN_ID> -R microsoft/openvmm \
-n aarch64-linux-unit-tests-junit-xml -D /tmp/junit-xml
python3 -c "
import xml.etree.ElementTree as ET, sys
for f in __import__('pathlib').Path(sys.argv[1]).rglob('*.xml'):
for tc in ET.parse(f).iter('testcase'):
fail = tc.find('failure')
if fail is None:
fail = tc.find('error')
if fail is not None:
print(f'FAIL: {tc.get(\"classname\",\"\")}::{tc.get(\"name\",\"\")}')
print(f' {fail.get(\"message\",\"\")[:200]}')
" /tmp/junit-xml
VMM test results are stored in artifacts named {platform}-vmm-tests-logs.
The known platforms are:
x64-windows-intel
x64-windows-intel-tdx
x64-windows-amd
x64-windows-amd-snp
x64-linux
aarch64-windows
gh run download <RUN_ID> -R microsoft/openvmm \
-n x64-windows-amd-snp-vmm-tests-logs -D /tmp/test-logs
4. Find failed tests
Each test gets its own directory inside the artifact. Look for petri.failed
marker files (passing tests have petri.passed instead):
find /tmp/test-logs -name "petri.failed"
The petri.failed file contains the error petri recorded for the test. The
test's real name is in petri.test, which petri writes before the test body
runs — so a directory with petri.test but no result marker is a test that
was killed or crashed before it could report.
5. Extract errors from petri.jsonl
The petri.jsonl file in each test directory is the primary structured log.
Each line is a JSON object with fields: timestamp, source, severity,
message. Filter for ERROR and WARN severity for a quick diagnosis:
python3 -c "
import json, sys
for line in open(sys.argv[1]):
try:
e = json.loads(line.strip())
if e.get('severity') in ('ERROR', 'WARN'):
print(f'[{e[\"severity\"]}] {e.get(\"source\",\"?\")}: {e.get(\"message\",\"\").strip()[:200]}')
except: pass
" /tmp/test-logs/<test-dir>/petri.jsonl
Artifact Contents
Unit test JUnit XML
Artifacts named {platform}-unit-tests-junit-xml contain JUnit XML files
with <testcase> elements. Failed tests have <failure> or <error>
children with message attributes describing the failure. These are the
primary artifacts for diagnosing unit test / cargo-nextest failures.
VMM test logs (petri)
Each test directory contains:
petri.jsonl — Structured JSON Lines log (primary file for investigation)
petri.log — Plain text version of the test log
petri.test — The test's name, written before the test body runs
petri.passed or petri.failed — Pass/fail marker; a failure marker holds the error
openhcl.log — OpenHCL serial console output, if the test exercised OpenHCL
hyperv.log — Hyper-V event log, if the test exercises the Hyper-V backend
openvmm.log — OpenVMM serial console output, if the test exercises the OpenVMM backend
guest.log, uefi.log — Guest OS serial output
- Sometimes:
screenshot_*.png — periodic screenshots of the guest
- Sometimes:
dumpfile.dmp
Viewing Results in Browser
Test results are uploaded to Azure Blob Storage and viewable at:
https://openvmm.dev/test-results/#/runs/<RUN_ID>_<RUN_ATTEMPT>
Logs are uploaded per re-run attempt, so the run key includes the attempt
number (1 for the first attempt). investigate_ci.py prints the correct
URL in its summary.
Common Failure Patterns
- Unit test failure: A
unit tests job failed. The script downloads
JUnit XML artifacts and shows the failing test names and messages. Common
causes: new test code that relies on OS capabilities not available in CI
(e.g. TAP devices, elevated permissions).
- Formatting / house-rules: The
quick check [fmt, clippy x64-linux] job
failed. No test artifacts will exist. Run cargo xtask fmt --fix locally
and check the job log for the specific rule that failed.
- Docs failure: A job in the separate "OpenVMM Docs PR" workflow failed —
either
build mdbook guide (broken Guide markdown/links; mdbook logs
[ERROR] (mdbook::...)) or build and check docs (rustdoc warnings, which
are denied in CI). Reproduce with cargo doc --no-deps -p <package>.
No test artifacts will exist.
- TripleFault: VM hit a fatal error during boot. Check
petri.jsonl for
Hyper-V Worker/Chipset errors. Often infrastructure-related, not caused by
the PR's code changes.
- Timeout: Test exceeded its time limit. Check if the VM booted at all.
- Guest assertion failure: Guest-side test failed. Check
guest.log.
- Build failure: No test artifacts will exist. Check the job log directly
with
gh run view <RUN_ID> --job <JOB_ID> --log.
Important API Note
The gh run view --json artifacts flag does not exist. To list artifacts
for a run, use the GitHub API directly:
gh api repos/microsoft/openvmm/actions/runs/<RUN_ID>/artifacts