On Wed, 2013-04-10 at 15:56 -0400, Simo Sorce wrote:
> On Wed, 2013-04-10 at 15:46 -0400, Rob Crittenden wrote:
> > Simo Sorce wrote:
> > > On Wed, 2013-04-10 at 15:35 -0400, Rob Crittenden wrote:
> > >> My problem with systemd is that it seems to make things more opaque.
> > >> It
> > >> is hard to do even simple things like an strace on process startup
> > >> because the shell is completely detached from execution. Which means
> > >> that your logging needs to be top-notch so when things go wrong we
> > >> can
> > >> look somewhere and say "aha, it's X"
> > >>
> > >
> > > Actually with systemd logging has become easier.
> > > All you need to do is spit stuff to stdout (of course you must avoid
> > > closing stdout if the daemon does so as they used to do to cope with
> > > sysv init madness :)
> > Yes, it is a mixed bag. Doesn't replace the sweetness of strace
> > /etc/init.d/service start to find startup permission issues. If your
> > daemon doesn't log that it is failing because it can't open some file
> > then you're in deep.
> OTOH strace /etc/init.d/service start will almost certainly run
> unconfined, so if the access issue is due to selinux policy you'll be
> left with a fist of flies anyway.
> That said I think you can change the systemd unit file to run a shell
> script with strace /usr/bin/myexec in it, have not tried that one yet
I'd love an option to start a socket activated daemon in valgrind. That
would be very useful...
Freeipa-devel mailing list