[smf-discuss] Reliably detecting restarts?

[Nicolas Williams]

On Tue, Mar 06, 2007 at 02:52:11PM -0700, Tom Whitten wrote:
> I'm going to need some help from the more experienced members of the SMF
> team on that question.  I don't know what happens if svc.configd cannot
> access the repository.  David or Liane, can you help out?

Using the source I find this in $SRC/cmd/svc/startd/fork.c:

                        switch (WEXITSTATUS(exitstatus)) {
                        ...
                        case CONFIGD_EXIT_DATABASE_BAD:
                                errstr = "database corrupt";
                                break;
                        ...
                        }

                        /*
                         * Remedial actions for various configd failures.
                         */
                        switch (WEXITSTATUS(exitstatus)) {
                        ...
                        default:
                                fork_sulogin(B_FALSE, "svc.configd exited "
                                    "with status %d (%s)\n",
                                    WEXITSTATUS(exitstatus), errstr);
                                goto retry;
                        }
                        ...


Losing the running repository seems to be rather Bad, and causes sulogin
to be started.  I imagine that the only reasonable way to recover at
that point is to reboot, so perhaps that's what svc.startd ought to do.

Nico
--