fix(startup): defer UME alarm sync so /health is ready quickly

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
oliver 2026-06-04 21:28:57 +08:00
parent d05b64b4c1
commit f562f50953
2 changed files with 20 additions and 5 deletions

View file

@ -815,9 +815,23 @@ def on_startup() -> None:
last_error=f"startup_thread_init_failed: {str(exc)[:180]}", last_error=f"startup_thread_init_failed: {str(exc)[:180]}",
) )
try: try:
_run_startup_alarm_sync_before_ws()
def _startup_alarm_sync_worker() -> None:
try:
_run_startup_alarm_sync_before_ws()
except Exception as exc:
_schedule_log.exception("startup: alarm sync before WSS failed: %s", exc)
complete_startup_alarm_sync_gate()
# Do not block HTTP /health on slow UME REST pull; WSS waits on startup_alarm_sync_gate.
t_startup_sync = threading.Thread(
target=_startup_alarm_sync_worker,
name="ume-startup-alarm-sync",
daemon=True,
)
t_startup_sync.start()
except Exception as exc: except Exception as exc:
_schedule_log.exception("startup: alarm sync before WSS failed: %s", exc) _schedule_log.exception("startup: alarm sync thread init failed: %s", exc)
complete_startup_alarm_sync_gate() complete_startup_alarm_sync_gate()
try: try:
if bool(getattr(settings, "ume_sync_alarms_current_enabled", True)): if bool(getattr(settings, "ume_sync_alarms_current_enabled", True)):

View file

@ -93,7 +93,7 @@ function Show-LogTail {
} }
function Test-NetxApiListening { function Test-NetxApiListening {
param([string]$HostName, [int]$LocalPort, [int]$WaitSec = 12) param([string]$HostName, [int]$LocalPort, [int]$WaitSec = 45)
$deadline = (Get-Date).AddSeconds($WaitSec) $deadline = (Get-Date).AddSeconds($WaitSec)
while ((Get-Date) -lt $deadline) { while ((Get-Date) -lt $deadline) {
try { try {
@ -135,8 +135,9 @@ if ($Background) {
Show-LogTail -Path $logFile Show-LogTail -Path $logFile
exit 1 exit 1
} }
if (-not (Test-NetxApiListening -HostName $BindHost -LocalPort $Port)) { $healthWaitSec = 45
Write-Host "[ERR] Port $Port not listening /health not OK within 12s (PID may still be starting or stuck on DB)." -ForegroundColor Red if (-not (Test-NetxApiListening -HostName $BindHost -LocalPort $Port -WaitSec $healthWaitSec)) {
Write-Host "[ERR] Port $Port not listening /health not OK within ${healthWaitSec}s (process may be stuck on DB or schema migration)." -ForegroundColor Red
Show-LogTail -Path $errFile Show-LogTail -Path $errFile
Show-LogTail -Path $logFile Show-LogTail -Path $logFile
exit 1 exit 1