Phase A4 of plans/implementation-plan-A-B.md. Implements the full job orchestrator (clone -> start -> wait -> probe -> build -> collect -> guaranteed cleanup) as a new commands/job.py click command, registered under python -m ci_orchestrator job. Backend selection goes through backends.load_backend(config) so Phase C can swap in remote drivers without touching the command. The legacy scripts/Invoke-CIJob.ps1 is replaced by a thin PS 5.1 shim that delegates to the Python CLI; tests/python/test_commands_job.py adds 13 cases covering Linux/Windows happy paths, override application, skip-artifact, and cleanup on every failure mode.
- Get-CIJobSummary.ps1: already present with full implementation (marked done)
- Get-BuildArtifacts.ps1: add -JobId/-Commit params; Write-ArtifactManifest writes
manifest.json (name, size, sha256, commit, jobId) after both Linux and Windows
artifact collection branches
- Invoke-CIJob.ps1: add -GuestCPU/-GuestMemoryMB params; insert post-clone VMX
override block (numvcpus/memsize rewrite) before vmrun start; pass -JobId/-Commit
to Get-BuildArtifacts calls
- Watch-RunnerHealth.ps1: add -GiteaUrl/-GiteaCredentialTarget params; optional
GET /api/v1/admin/runners check in Running branch warns if 0 online runners
- Install-CIToolchain-WinBuild2025.ps1: fill SHA256 hashes for Python 3.13.3,
7-Zip 26.01, Node.js 22.14.0, dotnet-install.ps1
- Measure-CIBenchmark.ps1: add deltaKB field (linked-clone disk footprint in KB)
measured post-clone; added to result object and summary Format-Table
Root cause: ssh-keygen -R prints 'Host X not found in F:\CI\State\known_hosts'
to stderr when the IP has no prior entry. In PS5.1 with ErrorActionPreference=Stop,
native stderr is converted to ErrorRecords before 2>$null can discard them,
making the command a terminating error caught by the job's catch block.
This caused all 4 concurrent burn-in jobs to fail after ~15s with:
Error: Host 192.168.79.XXX not found in F:\CI\State\known_hosts
Additionally, 4 concurrent SSH connections writing to the same known_hosts file
with no advisory locking risk file corruption under concurrent load.
Fix: set SshKnownHostsFile default to '' in Invoke-CIJob, Invoke-RemoteBuild,
and Get-BuildArtifacts. When KnownHostsFile='' _Transport.psm1 uses
StrictHostKeyChecking=no + UserKnownHostsFile=NUL (permissive mode), consistent
with Wait-VMReady and Prepare-LinuxBuild2404.ps1. Ephemeral CI VMs on a
local NAT network have no meaningful MITM risk.
Remove the ssh-keygen purge block from Invoke-CIJob.ps1 (no longer needed).
AGENTS.md: document PS5.1 native stderr / 2>$null interaction (error #12).
Each Linux BuildVM clone gets a new SSH host key for the same
NAT IP, causing 'host key has changed' errors with accept-new mode.
After IP detection, ssh-keygen -R removes the old entry from the
CI-local known_hosts file (F:\CI\State\known_hosts by default)
before Phase 4/5 make any SSH connection. The file is created if
it does not exist yet. Windows jobs are unaffected.
PROBLEM: Invoke-RemoteBuild.ps1 was creating a placeholder artifact when the build
produced no dist/ directory. In production this hides build defects (build exits 0
but produces nothing = silent failure treated as success).
SOLUTION: Add explicit -SkipArtifact switch to the call chain:
Invoke-RemoteBuild.ps1:
- Add -SkipArtifact switch. When set: skips packaging entirely (no zip created).
- When absent (production default): missing GuestArtifactSource is a hard error
with a clear message pointing to -SkipArtifact as the opt-in for no-output builds.
Invoke-CIJob.ps1:
- Add -SkipArtifact switch. Passes it to Invoke-RemoteBuild.ps1.
- Phase 6 (Get-BuildArtifacts) is entirely skipped when set; logs 'skipped' event.
Test-CapacityBurnIn.ps1:
- Add -SkipArtifact switch. Adds '-SkipArtifact' flag to each child job argList.
Start-BurnInTest.ps1:
- Always passes -SkipArtifact by default (burn-in uses delay commands, no output).
- Add -NoSkipArtifact override switch for runs that do produce real artifacts.
- Remove -NoSkipArtifact escape from wrapper: default burn-in never needs artifacts.
Add -UseGitClone switch to Invoke-CIJob.ps1 + conditional logic:
Changes:
1. Add -UseGitClone [switch] param + docstring (§3.3 mode)
2. Modify Phase 1 to skip host clone when -UseGitClone active
3. Phase 5: Pass CloneUrl/CloneBranch/CloneCommit instead of HostSourceDir
when -UseGitClone enabled
4. Finally: Skip host cleanup when -UseGitClone (dir never created)
When -UseGitClone is active:
- Phase 1 (host clone) skipped — clone happens in VM instead
- Phase 5 passes git repo parameters to Invoke-RemoteBuild.ps1
- PAT (if needed) read in Invoke-RemoteBuild session (§1.5)
Prerequisite: Git for Windows + 7-Zip in template (§6.6, now complete)
Work in progress: Invoke-RemoteBuild.ps1 modifications pending (source transfer
conditional + guest-side git clone logic).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
§4.1 — Structured JSONL log (Invoke-CIJob.ps1)
Each job now emits a parallel invoke-ci.jsonl alongside the transcript.
Write-JobEvent appends one JSON line per phase transition (ts, jobId,
phase, status, data). Events emitted at: job.start, phase1.clone-repo
start/success, phase2-3b.vm-start start/success, phase4.wait-ready
start/success, phase5.build start/success, phase6.artifacts start/success,
job.success/failure (with elapsedSec and error string).
Silent no-op if $jsonLog is null (log dir setup failed). Errors in
Write-JobEvent are swallowed — logging never breaks the build.
Enables post-hoc per-phase duration analysis with jq or ConvertFrom-Json.
§4.3 — Disk space alert (scripts/Watch-DiskSpace.ps1)
Checks drive free space every 15 min (scheduled by Register-CIScheduledTasks).
Below MinFreeGB (default 50 GB): writes Warning to Windows Application Event Log
(source CI-DiskAlert, EventId 1001) and exits 1 so Task Scheduler flags the run.
Optional -WebhookUrl for Discord/Gitea webhook notification.
Register-CIScheduledTasks.ps1 updated with Task 3: CI-DiskSpaceAlert (every 15 min).
§4.4 — Incident runbook (docs/RUNBOOK.md)
Four scenarios with symptom / triage commands / fix / escalation:
1. Runner offline in Gitea UI
2. All builds fail in Phase 2 (clone/start/IP)
3. Builds are slow (per-phase JSONL analysis, disk/CPU/NAT checks)
4. Template VMX corrupt after host crash (lock removal + backup restore)
Quick-reference table at the end.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
§2.1 — IP race with capacity>1 (Invoke-CIJob.ps1)
Phases 2-3b (clone+start+IP) now run inside a file-based exclusive mutex
(F:\CI\State\vm-start.lock via FileShare.None). Only one job at a time is
in the VM-start phase, preventing concurrent vmrun starts from racing for
the same DHCP lease. After IP is confirmed unique it is written to an IP
lease file (F:\CI\State\ip-leases\<IP>.lease). Lock is released immediately
so build phases run fully in parallel. Lease released in finally block.
Collision detection: if a lease file already exists for the detected IP,
the job fails fast with a clear error rather than silently sharing a WinRM target.
§2.5 — Versioned snapshot (Invoke-CIJob.ps1 + runner/config.yaml)
$SnapshotName now defaults to GITEA_CI_SNAPSHOT_NAME env var, falling back
to 'BaseClean'. config.yaml documents the commented-out variable for snapshot
refresh workflow (BaseClean_<yyyyMMdd> naming convention).
§2.3 — Artifact/log retention (scripts/Invoke-RetentionPolicy.ps1)
Purges per-job subdirs under F:\CI\Artifacts and F:\CI\Logs older than
RetentionDays (default 30). Switches to AggressiveRetentionDays (default 7)
when F: free space drops below MinFreeGB (default 50 GB). Also removes stale
IP lease files older than 12 hours (defensive cleanup for crashed jobs).
SupportsShouldProcess (-WhatIf) for dry-run.
§2.2 — Scheduled task registration (scripts/Register-CIScheduledTasks.ps1)
Registers two tasks under \CI\ task folder as SYSTEM/HighestPrivilege:
CI-CleanupOrphans — every 6h + AtStartup, -MaxAgeHours 4
CI-RetentionPolicy — daily 03:00 (+30min random) + AtStartup (+15min random)
Idempotent (-Force), SupportsShouldProcess (-WhatIf), validates script paths.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Previous fix removed --depth 1 entirely when $Commit was set, causing
full clones on every workflow run (github.sha is always non-empty).
Better approach: keep --depth 1 (fast path), detect checkout failure,
then fetch --unshallow and retry. Only pays full-clone cost when the
requested commit is actually older than HEAD.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Invoke-RemoteBuild.ps1:
- Custom build command: replace buffered collect-then-print with
Write-Host streaming inside Invoke-Command (WinRM forwards Write-Host
in real-time to the host console)
- Set PYTHONUNBUFFERED=1 in the remote session to prevent Python from
holding lines in its internal buffer
- Same streaming approach for dotnet restore and dotnet build paths
- Return value is now just int exit code instead of hashtable
Invoke-CIJob.ps1:
- Add -LogDir parameter (default F:\CI\Logs)
- Add -LogRetentionDays parameter (default 30)
- Start-Transcript at job start -> F:\CI\Logs\<JobId>\invoke-ci.log
- Stop-Transcript in finally block (guaranteed execution)
- Auto-purge log directories older than LogRetentionDays after each job
Simone
Simone