On a freshly installed hardened/xen system, sysctl-ng-3.0.4 will not start at boot, but will do when trying manually. Further tests shows that it simply refuses to start the first time you try (either from the default runlevel or manually invoking the init.d script via commandline) but will start correctly the second time you try. This gets repeated after every reboot. Profile hardened/linux/amd64/10.0, kernel 2.6.18.8 from Xen-3.4.2 package. Example output with syslog-ng in default runlevel: INIT: Entering runlevel: 3 * Starting syslog-ng .../usr/sbin/syslog-ng already running. * Failed to start syslog-ng [ !! ] If I remove it from the 'default' runlevel, reboot, and login as root, I can do the following: srv7 ~ # ps aux | grep syslog | grep -v grep srv7 ~ # /etc/init.d/syslog-ng status * status: stopped srv7 ~ # /etc/init.d/syslog-ng start * Starting syslog-ng .../usr/sbin/syslog-ng already running. * Failed to start syslog-ng [ !! ] srv7 ~ # /etc/init.d/syslog-ng start * Starting syslog-ng ... [ ok ] srv7 ~ # /etc/init.d/syslog-ng status * status: started upgrading to (currently masked for testing) syslog-ng-3.0.5 seems to fix the problem, probably due to the change made for bug #295394 I don't really know who is thinking syslog-ng is already started after a reboot (start-stop-daemon? it uses no pid file for syslog-3.0.4...) and why.
syslog-ng in portage is now at 3.0.6 stable on all archs, I can't reproduce the original issue any more.