* Dan Fayette <fayette at rocketmail.com> [2007-06-28 11:07]: > I've been doing a LOT of research on SMF and created several manifests and am > very happy with how all the pieces work and the additional functionality that > is now available. The one question that I have been unable to find an answer > to is the reason for this post. > > Before I list my question please understand that I have seen a lot of posts > from people asking why you would want to do this, or stating how stupid the > idea/concept is so please try to keep an open mind. The service I am using > is not the intended service that I want the behavior for it is just a good > service for testing the behavior without corrupting data. > > Basically I have a network service (stunnel) that I want to: > 1. start at boot time. > 2. Not try to restart if the process dies (IE: pkill -9 stunnel) > 3. Change the status from online to something else (IE: maintenance) > > Part 1 is not a problem. > Part 2 is doable if I use: > <property_group name='startd' type='framework'> > <propval name='duration' type='astring' value='transient' /> > </property_group> > > Part 3 is the tricky part. ( At least for me ) > > I saw a post > (http://www.opensolaris.org/jive/thread.jspa?messageID=114094𛶮) where > they wanted the same behavior for xntpd and the fix was to write C+ code. > That post was back in 2005 and I'm hoping that there is either a better way > to do it now or that it is at least being looked at as an option to add to > the SMF framework. > > With RC scripts the service will die and not try to restart itself but there > is also nothing that reports the service as being "online." With SMF I don't > want the service to report that it is online when it is not. > > If this has been discussed somewhere else I apologize for the repeated post > but I have done a lot of searching and have not been able to find a clear cut > answer to this problem. I hope that I'm not the only Solaris Admin in the > world that wants to do this but if I am please tell me what other options I > have. (Other than sticking with RC Scripts). Can you keep stunnel in the foreground? If so, you might be able to do something like
"stunnel ARGS || exit $SMF_EXIT_ERR_FATAL" as the start method. If not, you'll need to write a little method in shell that runs stunnel, and then does a wait, and then exits with a status with precise meaning to smf(5). In summary: if you can have your service fail with one of the SMF_EXIT_ERR_* values, then svc.startd(1M) won't attempt to restart it. - Stephen -- sch at sun.com http://blogs.sun.com/sch/
