On Thu, 11 Feb 2016 21:05:26 +0000 Martin Gregorie wrote: > On Thu, 2016-02-11 at 19:09 +0000, RW wrote: > > On Thu, 11 Feb 2016 12:02:13 +0000 > > Martin Gregorie wrote: > > > > > Every so often I see the following in my nightly logwatch rep > > > > > > Child-related errors > > > spamd: cannot send SIGINT to child process [___]: No such > > > process:
> > Is it literally "[___]" in the log? > > > No, that looks like logwatch's doing because, in maillog the child's > PID is inside the square brackets. > > This only happens during a shutdown: Yes, the warning comes from the SIGTERM/SIGINT handler, in the main spamd process, when it tries to kill all the children. Child processes are reaped in the SIGCHLD handler, but the PIDs are removed from the list later in the select loop, so there is a window where the SIGINT handler could find such a PID. > Feb 7 00:29:22 zoogz spamd[9347]: spamd: server killed by SIGTERM, > shutting down > Feb 7 00:29:22 zoogz spamd[9347]: spamd: cannot send SIGINT to child > process [9352]: No such process > Feb 7 00:29:22 zoogz spamd[9347]: spamd: cannot send SIGINT to child > process [9353]: No such process I do find it a bit strange that in both examples there are two child processes involved. I tried to reproduce this, but only saw one at a time.
