Slightly uncomfortable just sleeping for one second without >>> rechecking. >>> >> Yes, I don't like the wait myself. However, in my testing, 1 second is >> sufficient so if mountd hasn't come up then there maybe a problem. In >> the case that mountd didn't start, the /usr/lib/servinfo will not return >> any port so we simply don't generate any rules for mountd. >> > > Fair enough. Presumably SMF will restart the service in the case that > mountd failed anyway. > Actually, SMF would place the service in maintenance if mountd failed to start in which case we would remove the service's rules.
>>> usr/src/cmd/ipf/svc/ipfd.c >>> >>> Lines 537-543: That looks like a syslog entry that hasn't been >>> finished. Is this just going to write "fmri: svc://network/rpc/bind" >>> or whatever to a logfile? Won't that be confusing? Why not just >>> do nothing? >>> >> Yes, it can be removed but may serve some debugging purposes. How about >> something like: >> >> /* >> * Service, instance, or pg deleted. >> */ >> syslog(LOG_DEBUG | LOG_DAEMON, "Deleted: %s", fmri); >> > > Both the comment and the log entry give the impression that ipfd deleted it. > I think you'd probably want to avoid that. > > Ah, I see your concern. How about being clearer that the received event was a deletion? Perhaps, something like: /* * If we get here, the received event is a deletion of a service, * instance or pg. */ syslog(LOG_DEBUG | LOG_DAEMON, "Event: %s deleted", fmri); Thanks, -tony
