Just getting back to this...sorry for the delay. On Thu, 2008-02-07 at 13:42 -0400, Renier Morales wrote:
> > > > > +### BEGIN INIT INFO > > > > > +# Provides: openhpid > > > > > +# Required-Start: $network $named $remote_fs $syslog > > > > > +# Required-Stop: $network $named $remote_fs $syslog > > > > > +# Default-Start: 2 3 4 5 > > > > > +# Default-Stop: 0 1 6 > > > > > +# Short-Description: Start OpenHPI daemon at boot time > > > > > +# Description: Enable OpenHPI service which is provided > > by openhpid. > > > > > +### END INIT INFO > > > > The reference I used for the "Required-Start" and "Required-Stop" > was: > > > > http://refspecs.freestandards.org/LSB_3.1.0/LSB-Core-generic/LSB- > > Core-generic/facilname.html > > > > For "$named", my reasoning was that some installations may want > > hostnames in openhpi.conf, and without $named, the daemon start-up > for > > those would fail. I would suggest erring on the side of safety, > because > > it can be quite hard for a user to figure out a problem such as > this, > > when a daemon won't start while booting, but will start later when > run > > by hand. An example of a plugin that can take a hostname is > ipmidirect, > > when using the rmcp protocol. > > $named is a DNS server, IIRC. The only case where not requiring it to > start would cause a problem is if the system where the openhpi daemon > was running also was its own DNS server, which is usually not the > case. But if it doesn't hurt, then might as well leave it. > My question would be, for those servers that don't need to run a DNS > server to resolve hostnames, will they be forced to run a DNS server > in order to run openhpi? Is that how the LSB init require's work? I exchanged e-mail with Petter Reinholdtsen, who has done a ton of work in this area. He suggested we use Should-Start and Should-stop in place of Required-Start and Required-Stop. In this way, we hint that we need those services to be completely functional, but that they are not required. > > For "$remote_fs", the above description says that some LSB > installations > > may have /usr as a remote filesystem. Since (I think) our libraries > go > > under /usr in many installations, don't we need this so that the > daemon > > can load the plugin shared library? > > Same question as with $named... Petter confirmed that this needs to be Required-Start, because we depend on /usr being mounted. This seems to be an unfortunate choice of names, but what it's really saying is that we can't start until remote filesystems are mounted. Of course, if there are no remote filesystems, we have nothing to wait for. His comments, verbatim: > The $remote_fs dependency (at least in Debian) specify a point in time > during boot when /usr/ and all remote file systems if they exist are > mounted. It is not (in Debian) depending on the existence of remote > file systems nor any network. It simply mean that the script will > start after mountnfs.sh during boot. In SUSE the behaviour is > similar. I would now suggest the following: ### BEGIN INIT INFO # Provides: openhpid # Required-Start: $network $remote_fs $syslog # Required-Stop: $network $remote_fs $syslog # Should-Start: $named # Should-Stop: $named # Default-Start: 2 3 4 5 # Default-Stop: 0 1 6 # Short-Description: Start OpenHPI daemon at boot time # Description: Enable OpenHPI service which is provided by openhpid. ### END INIT INFO Please respond if you disagree. In the interest of being proactive, I'll check in this change, but we can certainly change it if we want to. Bryan Sutula ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Openhpi-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/openhpi-devel
