Retry on non-ServiceError exceptions in ARM monitor

This commit is contained in:
Gan, Jimmy
2026-02-19 21:59:20 +08:00
parent 607432bb50
commit 043924b747
+3
View File
@@ -69,6 +69,9 @@ while True:
print(f"Unexpected error: {e.message}", flush=True) print(f"Unexpected error: {e.message}", flush=True)
send_email("OCI ARM Monitor Error", str(e.message)) send_email("OCI ARM Monitor Error", str(e.message))
sys.exit(1) sys.exit(1)
except Exception as e:
attempts += 1
print(f"{time.strftime('%Y-%m-%d %H:%M:%S', time.gmtime(time.time() + 8*3600))} {e}. Retrying in {INTERVAL}s...", flush=True)
now = time.time() now = time.time()
if now - last_report >= WEEKLY_REPORT_INTERVAL: if now - last_report >= WEEKLY_REPORT_INTERVAL:
uptime_days = int((now - start_time) / 86400) uptime_days = int((now - start_time) / 86400)