On Thu, 21 Oct 2010, Colin McCabe wrote: > On Thu, Oct 21, 2010 at 4:41 PM, Sage Weil <sage@xxxxxxxxxxxx> wrote: > > I wonder if a better strategy would be to _not_ delete the objects, but to > > create a placeholder, and mark it such that any attempts to read it return > > EIO or ESTALE or something along those lines. That would let an > > application know when data is gone instead of 'silently' (well, at the > > behest of a desperate administrator) losing the data. Things like remove > > and replace would succeed, but reads would not. Stale objects could then > > always be removed on a per-object basis. > > It sounds promising. I wonder what the metadata servers should do if > they get an ESTALE? They should probably pass it back to the user, in one form or another. For now at least it's can be treated the same as EIO or ENOENT. > Others using the object store would probably appreciate the > flexibility of being able to read other objects which didn't get lost. > > We probably want to make the "ceph health" command comment about lost > objects if they exist. Right. The pg_stat_t structure can be extended to include lost objects (if it doesn't already), so that it can be included in the pool and system totals on the monitors. sage
