Nicolas Williams wrote: > Hmmm, how does svc.startd recover FMRI status when it dies and is > restarted?
From the repository. We don't need to decorate contracts to recover from svc.startd death. We've been recovering from svc.startd (and init, and svc.configd, ...) death since very early in the initial SMF project, and do so by storing the required information to recover in the repository. It's part of the reason we needed a transactional repository, even for non-persistant data (e.g. the service states, and the fmri-to-contract mapping). I've no real desire to revisit that fundamental design decision. liane
