Quoth Keith M Wesolowski on Wed, Jan 07, 2009 at 01:39:19PM -0800: > Are there plans to implement the degraded state? At present any call > to smf_degrade_instance() results in startd informing us that > ADMIN_DEGRADED is not implemented.
I know of no such plans. > If not, what work is required to implement this feature correctly? The first step is making startd/restarter.c respond to degrade requests by changing the service state. I believe most of the code in startd/graph.c should know that degraded is equivalent to online for most purposes, but there could be some problems. Then svcs -x's degraded handling may need to be fixed or implemented. > An example use case of interest is having ldap_cachemgr place itself > into the degraded state when none of the configured LDAP servers is > reachable (Compare with the present behaviour of having the start > method time out three times and then entering the maintenance state, > forcing local intervention to repair a remotely-induced condition). If I recall correctly, the way we want that to be modeled is by the ldap service claiming to be online as soon as it can serve requests, even if the replies are "no servers responded yet". Of course, it may be that the name service client / ldap client daemon protocol doesn't contain a such a response, in which case degraded may be the only alternative. David
