> -----Original Message----- > From: Liu Yuan [mailto:namei.unix at gmail.com] > Sent: Friday, December 30, 2011 3:06 PM > To: Peng, Tao > Cc: morita.kazutaka at lab.ntt.co.jp; sheepdog at lists.wpkg.org > Subject: Re: [Sheepdog] [PATCH v4 11/12] sheep: add cluster snapshot/restore support > > On 12/30/2011 02:50 PM, tao.peng at emc.com wrote: > > > How does the above become a cluster wide snapshot? From what I read, farm.snapshot only ensures > single node's dirty objects to be snapshotted. Or am I missing something? > > > > > Yes, snapshot operation of the Farm backend store is on node basis. We > just send an cluster-wide message that request for snapshot operations > on each node. > > Currently, Farm doesn't support *live* snapshot, so it safe for us to do > a cluster-wide snapshot without on-the-fly IO requests. And those > snapshot objects are distributed on each backend store. > I see. Thanks for the explanation. Cheers, Tao |