BTW, here's the patch I propose: *** Daemon.py 14 Aug 2002 22:12:52 -0000 1.11 --- Daemon.py 5 Oct 2002 08:14:51 -0000 *************** *** 17,22 **** --- 17,28 ---- import zLOG from SignalPasser import SignalPasser + # If zdaemon finds that it is continuously respawning more than 10 + # times in 2 minutes, it will assume that there is an error, log a + # PANIC level message, and exit. + RESPAWN_TIME = 120 # 2 minutes + RESPAWN_LIMIT = 10 # 10 times + pyth = sys.executable class DieNow(Exception): *************** *** 32,38 **** --- 38,54 ---- if not os.environ.has_key('Z_DEBUG_MODE'): detach() # detach from the controlling terminal + starttimes = [] # for RESPAWN_LIMIT while 1: + # Give up if respawning too often + starttimes.append(time.time()) + if len(starttimes) > RESPAWN_LIMIT: + del starttimes[0] + if starttimes[-1] - starttimes[0] < RESPAWN_TIME: + pstamp('Respawning more than %d times in %d seconds. Quit.' % + (RESPAWN_LIMIT, RESPAWN_TIME), zLOG.PANIC) + sys.exit(1) + del starttimes[0] try: pid = os.fork() if pid: While we're improving zdaemon, I have two other suggestions: - Use grown-up language in log messages rather than "Aiieee!" and "Houston, we have forked" - Why does zdaemon add "zdaemon: <timestamp>: " to its log messages? That info is already added by zLOG. A typical set of zdaemon log messages now looks like this: ------ 2002-10-05T04:23:31 INFO(0) zdaemon zdaemon: Sat Oct 5 04:23:31 2002: Houston, we have forked: pid 2915 ------ 2002-10-05T04:24:12 ERROR(200) zdaemon zdaemon: Sat Oct 5 04:24:12 2002: Aiieee! Process 2915 terminated normally, exit status: 1 ------ while it could look like this: ------ 2002-10-05T04:23:31 INFO(0) zdaemon Started subprocess: pid 2915 ------ 2002-10-05T04:24:12 ERROR(200) zdaemon Process 2915 terminated normally, exit status: 1 ------ --Guido van Rossum (home page: http://www.python.org/~guido/)