Okay, I will just poke the upstream about it, seems that CentOS/RHEL package have exactly same problem so no point changing it just for Debian.
On Thu, 24 Jan 2019 at 18:27, Michael Stapelberg <stapelb...@debian.org> wrote: > > Ah, why didn’t you open with that suggestion? :) > > https://manpages.debian.org/stretch/systemd/systemd.service.5.en.html#OPTIONS > outlines that switching to Type=simple (which is the consequence of your -f > suggestion) means we won’t have start-up failure propagation anymore, and > reverse-dependencies of freeradius will not be able to reliably sequence > themselves. > > Now, I’m not sure whether these downsides are actually relevant in practice, > as I don’t know much about the larger freeradius ecosystem. Maybe there are > no communication channels, and no reverse dependencies of note? I’m not sure. > > If you could confirm with upstream that they are blessing use of -f and > Type=simple as the default in Debian, I can make the change. > > Thanks, > > On Thu, Jan 24, 2019 at 4:49 PM Mariusz Gronczewski <xani...@gmail.com> wrote: >> >> Well, there is, adding -f option by default means it will always run >> in foreground, regardless of whether -X is used or not >> >> On Thu, 24 Jan 2019 at 15:58, Michael Stapelberg <stapelb...@debian.org> >> wrote: >> > >> > The -X option is special in that it changes the way freeradius starts up. >> > >> > It’s expected that if your actions break the contract with the init system >> > (in this case by specifying -X), you’re responsible to rectify that. >> > >> > If there was a simple way to make the package work in any/all cases, it’d >> > be up for it, but I’m not aware of one. Hence my question for what your >> > suggestion is — I just don’t see a way out of this situation. >> > >> > My personal recommendation is to not use /etc/default, but rather >> > systemctl edit to do any overrides, but that’s not Debian’s official line. >> > >> > On Thu, Jan 24, 2019 at 3:37 PM Mariusz Gronczewski <xani...@gmail.com> >> > wrote: >> >> >> >> In our case it was enough to add dropin changing execstart to include >> >> -f and type to simple: >> >> >> >> ExecStart=/usr/sbin/freeradius -f $FREERADIUS_OPTIONS >> >> Type=simple >> >> >> >> What do you mean by "it is expected" ? Currently enabling debug (by >> >> setting FREERADIUS_OPTIONS="-X") cause it to enter restart loop, >> >> surely that isn't an expected behaviour ? >> >> >> >> On Thu, 24 Jan 2019 at 13:52, Michael Stapelberg <stapelb...@debian.org> >> >> wrote: >> >> > >> >> > Yes, this is expected. What change are you suggesting? >> >> > >> >> > On Thu, Jan 24, 2019 at 1:45 PM Mariusz Gronczewski <xani...@gmail.com> >> >> > wrote: >> >> >> >> >> >> Package: freeradius >> >> >> Version: 3.0.12+dfsg-5+deb9u1 >> >> >> Severity: normal >> >> >> >> >> >> Currently the type of systemd service is forking. >> >> >> >> >> >> Adding debug to cmdline causes freeradius to run in foreground (and >> >> >> dump debug >> >> >> to stdout), which means systemd timeouts on starting service because >> >> >> it assumes >> >> >> it will fork. >> >> >> >> >> >> Changing type to simple, and adding -f (run in foreground) option in >> >> >> unit file >> >> >> fixes that >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> -- >> >> >> Mariusz Gronczewski (XANi) <xani...@gmail.com> >> >> >> GnuPG: 0xEA8ACE64 >> >> >> >> >> >> _______________________________________________ >> >> >> Pkg-freeradius-maintainers mailing list >> >> >> pkg-freeradius-maintain...@alioth-lists.debian.net >> >> >> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-freeradius-maintainers >> >> > >> >> > >> >> > >> >> > -- >> >> > Best regards, >> >> > Michael >> >> >> >> >> >> >> >> -- >> >> Mariusz Gronczewski (XANi) <xani...@gmail.com> >> >> GnuPG: 0xEA8ACE64 >> > >> > >> > >> > -- >> > Best regards, >> > Michael >> >> >> >> -- >> Mariusz Gronczewski (XANi) <xani...@gmail.com> >> GnuPG: 0xEA8ACE64 > > > > -- > Best regards, > Michael -- Mariusz Gronczewski (XANi) <xani...@gmail.com> GnuPG: 0xEA8ACE64
