Hi Irek, Got it, Thanks :) ? idzzy On August 22, 2014 at 6:17:52 PM, Irek Fasikhov (malmyzh at gmail.com) wrote: node1: 4[TB], node2: 4[TB], node3: 4[TB] :) 22 ???. 2014 ?. 12:53 ???????????? "idzzy" <idezebi at gmail.com> ???????: Hi Irek, Understood. Let me ask about only this. > No, it's for the entire cluster. Is this meant that total disk amount size of all nodes is over than 11.8 TB? e.g ?node1: 4[TB], node2: 4[TB], node3: 4[TB] not each node. e.g ?node1: 11.8[TB], node2: 11.8[TB], node3:11.8 [TB] Thank you. On August 22, 2014 at 5:06:02 PM, Irek Fasikhov (malmyzh at gmail.com) wrote: I recommend you use replication, because radosgw uses asynchronous replication. Yes divided by nearfull ratio. No, it's for the entire cluster. 2014-08-22 11:51 GMT+04:00 idzzy <idezebi at gmail.com>: Hi, If not use replication, Is it only to divide by nearfull_ratio? (does only radosgw support replication?) 10T/0.85 = 11.8 TB of each node? # ceph pg dump | egrep "full_ratio|nearfulll_ratio" full_ratio 0.95 nearfull_ratio 0.85 Sorry I?m not familiar with ceph architecture. Thanks for the reply. ? idzzy On August 22, 2014 at 3:53:21 PM, Irek Fasikhov (malmyzh at gmail.com) wrote: Hi. 10?B*2/0.85 ~= 24 TB?with two replications,?total volume for the raw data. -- ? ?????????, ??????? ???? ??????????? ???.: +79229045757 -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.ceph.com/pipermail/ceph-users-ceph.com/attachments/20140822/8052b71f/attachment.htm>