On Wed, 9 Mar 2016, Vero Estrada Galinanes wrote: > Hi Sage, > Thanks for your answers > > > > In Ceph clusters disks are generally never powered down. It sounds like > > Schwarz et al were studying a different type of storage system. In our > > case, it's all about how quickly you discover a defect and repair around > > it. > > Indeed, he was studying archival storage. I’m interested in both system > types. > > > We do let you change the scrub intervals on a per-pool basis as well (see > > pool_opts in osd/osd_types.h), so you can adjust the policy based on both > > the pool type (replicated vs erasure coded) or the importance of the data > > it contains. > > > > That setting is exactly what I was expecting ceph should have. Do you > have any study done about choosing scrub intervals policy on > replicated/erasured pools? Nope! I think this would be very interesting for users. sage
