Hi, On Wed, Mar 9, 2016 at 9:21 AM, Veronica Estrada Galinanes <veg@xxxxxxxx> wrote: > From ceph website: "Deep scrubbing (weekly) reads the data and uses > checksums to ensure data integrity.” > > 1. Why do you use "weekly" deep scrubbing? How many times a year one disk > is scrubbed on average in a Ceph cluster? Schwartz* used three times per > year. “Interestingly, we did not observe a further decrease in data loss as > we increased the scrubbing frequency from three to eleven times per year; > rather, a slight increase in data loss was noticed. This phenomenon comes > from the POH (power-on-hour) effect on drive reliability. Aggressive > scrubbing requires more power cycles, adversely affecting drive reliability. > " Is power cycling relevant for modern enterprise drive? APM is disabled by default on all recent drives I've used. And their model is not realistic for Ceph: "We varied the access rate between once per 102 hours (about 4 days) to once per 8103 hours." With xfs/ext4 OSDs, if a random bit flip corrupts the master copy of an object, Ceph will happily serve that corrupted data until deep scrub notices the inconsistency between master and replicas. For this reason alone I prefer to deep scrub as often as possible (while keeping it transparent to the users). > 2. Do you have different scrubbing policies according to the redundancy > method? For example, more scrubs when using replication rather than erasure > coding? Scrubbing is configured with the parameters here: http://docs.ceph.com/docs/master/rados/configuration/osd-config-ref/#scrubbing -- Dan > > *Schwarz, Thomas JE, et al. "Disk scrubbing in large archival storage > systems." *Modeling, Analysis, and Simulation of Computer and > Telecommunications Systems, 2004.(MASCOTS 2004). Proceedings. The IEEE > Computer Society's 12th Annual International Symposium on*. IEEE, 2004. > > Thanks, > > -- > > Vero EG > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
