tis 2009-08-04 klockan 14:41 +0200 skrev Zeller, Jan: > signal 6 like mentioned in > > Aug 4 10:27:27 mybox squid[11814]: Squid Parent: child process 13220 exited due to signal 6 with status 0 > Aug 4 10:27:30 mybox squid[11814]: Squid Parent: child process 14441 started > Aug 4 10:27:30 mybox squid[14441]: Starting Squid Cache version 3.1.0.12-20090 > > kills the whole http service ? Yes. > So I probably send an SIGUSR1 during my tests... Probably. That's graceful. > Anyway the service interruption should be very short while sending signal 6 ...? according to the log snippet approximately 3 seconds... Depends on your configuration and definition of fast. But yes, for most it's quite fast. > What's about the keep-alive thing in http ? http://en.wikipedia.org/wiki/HTTP_persistent_connection > Sorry I am really confused about how loadbalancers e.g. linuxvirtualserver.org are dealing with such a short service interruption. lvs as such doesn't care as it just shuffles packets, but for example ldirectord or heartbeat MAY notice if the restart of Squid happened just as it is probing the service. Regards Henrik
