On Mon, 2009-11-16 at 16:01 -0500, Stephen Gallagher wrote: > > > When offline, we check at regular intervals to see if the backend > has come back online. This means that every n seconds (the online > check interval), NSS requests may lag as we attempt to go out to > the remote server and update the cache. > > With this patch, we will now always return the cached data while > we are offline and queue up an out-of-band cache refresh to > determine whether the data provider is back online.
Uhmm, unfortunately I have to nack this one. The problem is that you can't trust what you receive to be authoritative for the future. For example, between the last call from the nss frontend and the new one the backend can have run an enumeration task that have discovered we are back online. In that case it would probably be advisable to do an online refresh, except that you don't know that the backend is actually online because last time you got an answer it was offline. We should probably always contact the backend and let the backend know we want a "quick" answer by adding a parameter in the sbus call. If we require a "quick" call and the backend is offline it will immediately return back an offline error and then schedule an out of band request to try to go online. Simo. -- Simo Sorce * Red Hat, Inc * New York _______________________________________________ sssd-devel mailing list sssd-devel@lists.fedorahosted.org https://fedorahosted.org/mailman/listinfo/sssd-devel
