On Fri, 2005-06-24 at 15:53 -0400, Chris Santerre wrote: > > I'm completely guessing out of the blue here, but is it a timing issue? Is > it trying to restart before the final child is able to quit? > Yes, sort of...
Chris Thielen wrote: > > Maybe try changing your SA_RESTART to "killall -HUP spamd". I think > spamd will correctly reload configuration files with a HUP signal > Oddly enough for some reason when I did a 'killall -HUP spamd' all the spamd processes were killed off rather than restarted! Secondly, the spamd man page mentions a warning about using HUP. It turns out that the problem is not with SA, or RDJ, as such, but with Fedora. To 'restart' SA, the fedora startup script issues a 'stop' then a 'start'. To help with this the 'stop' uses some script functions in /etc/init.d/functions. In particular it uses 'killproc'. That function tries to locate the pids of the spamd processes. It first looks for a pid file, /var/run/spamd.pid, but that doesn't exist. So it then calls the 'pidof' command. This returns a list of the pids. However, it seems that the child processes are listed first, so the last pid is the parent one. It seems that as the children are being killed off, the parent sees this and restarts a child! Hence the overall 'restart' fails. I have inserted 2 lines into 'killproc' to reverse the pid order. Now the parent pid is seen first and killed off first. Testing this, 'restart' now works fine every time. I'll submit this as a bug to Fedora Core to see what they say. Needless to say, this problem may be particular to Fedora, other unix/linuxes may handle restarts and/or pids differently, and so not see this problem. Many thanks for all the replies. Now we have sorted this 'bug' out, I'll see about automating RDJ :-) John. -- --------------------------------------------------------------- John Horne, University of Plymouth, UK Tel: +44 (0)1752 233914 E-mail: [EMAIL PROTECTED] Fax: +44 (0)1752 233839
