ci: make the windev CI lock exclusive, stop the scheduled double run, isolate test secrets stores (mxaccessgw#150)
ci / nightly-windev (push) Has been skipped
ci / windows-x86 (push) Successful in 1m34s
ci / java (push) Successful in 2m37s
ci / portable (push) Failing after 5m0s

Diagnosis of the retained Windows-nightly failures (119 scheduled nightlies, 28 red; job logs from
the Gitea API, C:\build\mxaccessgw-ci\ci.log and the windev System log, read-only):

- Checkout (#148, run 827, 2026-09-22): the windev "atomic mkdir" lock was not exclusive.
  New-Item -ItemType Directory checks for the directory and then calls CreateDirectory, which
  succeeds on an existing directory; a pwsh race harness had both contenders take it in 1900 of
  2000 simultaneous rounds. A scheduled run starts windows-x86 and nightly-windev in the same
  second, both bootstraps got in, and the nightly's checkout died on .git/index.lock.
  Fix: the bootstrap (and a standalone windev-worker-ci.ps1) holds an exclusive FileShare.None
  handle on C:\build\mxaccessgw-ci.ci-lock instead (0 of 1000 double acquisitions in the same
  harness). The OS drops it if the holder dies, so a killed run leaves no stale lock, and a holder
  removes a leftover .git\index.lock, which can only be debris from a killed git.
- windows-x86 no longer runs on the schedule: nightly-windev's live mode already runs the same
  Worker build + tests first, so the scheduled copy only doubled the load and created the race.
- Host resources (#128/#129/#131-#135/#138, 2026-07-24..08-03): windev's System log has
  Resource-Exhaustion-Detector event 2004 (low virtual memory) every few minutes from 2026-07-23
  until an unexpected shutdown/hard reset on 2026-08-03; OOM in MSBuild/csc/protoc,
  csc 0xC0000142, PowerShell dying before its first log line and SSH kex resets all fall in that
  window and stop after the reset. Host-side; windev-worker-ci.ps1 now logs free commit, physical
  memory, disk and testhost count at start and on failure so the next one is correlatable.
- SSH "No route to host" (#144, 2026-09-16): windev was down (unexpected shutdown 00:58, back 05:50).
- Dashboard host build (#143, run 817): two host-building test classes failed at the same instant
  in SqliteSecretsStoreMigrator with "cannot start a transaction within a transaction" -- parallel
  classes migrating the one per-process secrets store. Tests now build the gateway through
  TestSupport/TestGatewayHost, which gives each host its own store; TestGatewayHostTests guards it.
- Gateway E2E "Worker pipe disconnected" (#142/#145/#146/#147) and handshaking-to-ready (#141) were
  the fake-worker partial-frame cancellation and readiness-race defects fixed in #153 (d779daf6):
  7 gateway-suite failures in 51 windev runs before it, 0 in 28 since. The Worker watchdog and
  burst-coalescing failures (#126/#127/#139) last failed before 50322bac / 7da52b65 (Aug 2026).
- 2026-10-15..20 was the NU1903 package-audit restore failure, fixed in fb20997.

Docs: GatewayTesting.md (lock, host-resource triage) and scripts/ci/README.md.

Claude-Session: https://claude.ai/code/session_01GT67AEHd2Acv1YPwGfXEoV
This commit is contained in:
Joseph Doherty
2026-11-09 13:35:54 -05:00
parent 09b3ecddf2
commit de86f8b80e
18 changed files with 234 additions and 65 deletions
+50 -16
View File
@@ -44,7 +44,7 @@ param(
Set-StrictMode -Version Latest
$ErrorActionPreference = 'Stop'
$lockDir = "$RepoRoot.lock"
$lockFile = "$RepoRoot.ci-lock"
$logFile = Join-Path $RepoRoot 'ci.log'
$lockTimeout = [TimeSpan]::FromMinutes(20)
@@ -67,37 +67,70 @@ function Invoke-Native {
}
}
# --- acquire the worktree lock (mkdir is atomic; retry until timeout) ---
# Logs the host's memory/commit/disk headroom and any leftover test processes, so a
# host-resource failure (OOM in csc/MSBuild/protoc, 0xC0000142 process-start failures, a
# PowerShell that dies before logging) can be correlated with the host's state at the time
# rather than reconstructed afterwards from the event log (mxaccessgw#150). Never fatal.
function Write-HostResources {
param([string]$When)
try {
$os = Get-CimInstance -ClassName Win32_OperatingSystem
$drive = Get-PSDrive -Name ($RepoRoot.Substring(0, 1)) -ErrorAction Stop
$testhosts = @(Get-Process -Name 'testhost*' -ErrorAction SilentlyContinue).Count
Write-Log ('host ({0}): commit free {1:N1} of {2:N1} GB, physical free {3:N1} of {4:N1} GB, {5}: free {6:N1} GB, testhost processes {7}, up since {8:yyyy-MM-dd HH:mm}' -f
$When,
($os.FreeVirtualMemory / 1MB), ($os.TotalVirtualMemorySize / 1MB),
($os.FreePhysicalMemory / 1MB), ($os.TotalVisibleMemorySize / 1MB),
$drive.Name, ($drive.Free / 1GB), $testhosts, $os.LastBootUpTime)
}
catch {
Write-Log "host ($When): resource probe failed: $($_.Exception.Message)"
}
}
# --- acquire the worktree lock ---
# The lock is an exclusive (FileShare.None) handle on "$RepoRoot.ci-lock". It replaced a
# `New-Item -ItemType Directory` lock that was not mutually exclusive (PowerShell checks for
# the directory and then calls CreateDirectory, which succeeds on an existing one), so two
# runs starting together both got in and collided on .git/index.lock (mxaccessgw#148/#150).
# The OS releases the handle if the holder is killed, so a dead run cannot leave it stale.
# run-windev-ci.sh's bootstrap already holds this lock around its pre-hand-off fetch/checkout
# (MXGW_CI_LOCK_HELD=1) so that stage can't race a concurrent run; re-use it rather than
# re-acquire (and leave releasing to the bootstrap). A standalone/manual run (degraded mode)
# has no such env and self-locks here as before.
# (MXGW_CI_LOCK_HELD=1); re-use it rather than re-acquire (and leave releasing to the
# bootstrap). A standalone/manual run (degraded mode) has no such env and self-locks here.
$deadline = (Get-Date).Add($lockTimeout)
$haveLock = $false
$lock = $null
if ($env:MXGW_CI_LOCK_HELD -eq '1') {
Write-Log "Worktree lock already held by bootstrap (mode=$Mode sha=$Sha)"
}
else {
Write-Log "Acquiring worktree lock $lockDir (mode=$Mode sha=$Sha)"
while (-not $haveLock) {
Write-Log "Acquiring worktree lock $lockFile (mode=$Mode sha=$Sha)"
while ($null -eq $lock) {
try {
New-Item -Path $lockDir -ItemType Directory -ErrorAction Stop | Out-Null
$haveLock = $true
$lock = [System.IO.File]::Open($lockFile, 'OpenOrCreate', 'ReadWrite', 'None')
}
catch {
if ((Get-Date) -ge $deadline) {
throw "Timed out after $($lockTimeout.TotalMinutes) min waiting for worktree lock $lockDir. A stuck run may have left it; if no CI job is active, remove it by hand."
throw "Timed out after $($lockTimeout.TotalMinutes) min waiting for worktree lock $lockFile ($($_.Exception.Message)). Another CI run is holding it."
}
Start-Sleep -Seconds 10
}
}
}
Write-HostResources -When 'start'
try {
Push-Location $RepoRoot
try {
# Re-fetch + re-checkout under the lock so a concurrent run's bootstrap checkout
# (done before we held the lock) cannot leave us on the wrong tree.
# Holding the lock proves no other CI git command is running in this clone, so a
# leftover index.lock can only be debris from a killed git; it would fail the checkout.
if (Test-Path '.git\index.lock') {
Write-Log 'Removing stale .git\index.lock left by a killed git'
Remove-Item -Force '.git\index.lock'
}
# Re-fetch + re-checkout under the lock so the tree is the requested SHA even when the
# bootstrap's checkout was skipped (standalone run) or left residue.
Invoke-Native 'git fetch origin' { git fetch --prune origin }
Invoke-Native "git checkout --force $Sha" { git checkout --force --detach $Sha }
# Drop any stray tracked/untracked residue from a prior run so the build is clean.
@@ -151,11 +184,12 @@ try {
}
catch {
Write-Log "FAILURE mode=$Mode sha=$Sha : $($_.Exception.Message)"
Write-HostResources -When 'failure'
throw
}
finally {
if ($haveLock) {
Remove-Item -Path $lockDir -Recurse -Force -ErrorAction SilentlyContinue
Write-Log "Released worktree lock $lockDir"
if ($null -ne $lock) {
$lock.Dispose()
Write-Log "Released worktree lock $lockFile"
}
}