Hi Yehuda&Craig Lewis, Thank you very much for your explanations, does the head object have some info about how to find the objects constituting the file, which is stripped? I am interested in the construction of the head object, but I can't see content of it. Thanks & Regards Li JiaMin -----邮件原件----- 发件人: ceph-users-bounces@xxxxxxxxxxxxxx [mailto:ceph-users-bounces@xxxxxxxxxxxxxx] 代表 Yehuda Sadeh 发送时间: 2014年3月20日 2:46 收件人: Craig Lewis 抄送: ceph-users@xxxxxxxxxxxxxx 主题: Re: 答复: why objects are still in .rgw.buckets after deleted The RGW objects are [for most cases] immutable. Therefore in order to provide read and write consistency we keep most of the data in immutable rados objects, and do the operations on the mutable 'head' part of the object atomically. So this allows us to do stuff like having one user read data of the object, while a second user overwrites the object and still the first user reads data that is consistent. When an object is removed we remove the head part and the rest is later on garbage collected. When it's rewritten, we write a new immutable 'tail' for and just replace the head at the end (then the old tail is garbage collected). Yehuda On Wed, Mar 19, 2014 at 11:31 AM, Craig Lewis <clewis@xxxxxxxxxxxxxxxxxx> wrote: > My understanding, from dealing with replication, is RadosGW is > copy-on-write. Overwriting an object is a delete and create, and the > old data gets garbage collected later. > > I'm guessing a bit, but that's what I believe from Greg's comment > about RGW > replication: > http://permalink.gmane.org/gmane.comp.file-systems.ceph.user/7434 > > > Craig Lewis > Senior Systems Engineer > Office +1.714.602.1309 > Email clewis@xxxxxxxxxxxxxxxxxx > > Central Desktop. Work together in ways you never thought possible. > Connect with us Website | Twitter | Facebook | LinkedIn | Blog > > On 3/19/14 03:58 , ljm李嘉敏 wrote: > > You are right, but I still don’t know why the objects in .rgw.buckets > are not overrided. > > If the object name is produced through ino and ono, why the same > file(bigger than 4M) have different result? > > > > Thanks & Regards > > Li JiaMin > > > > 发件人: ceph-users-bounces@xxxxxxxxxxxxxx > [mailto:ceph-users-bounces@xxxxxxxxxxxxxx] 代表 Craig Lewis > 发 送时间: 2014年3月19日 9:00 > 收件人: ceph-users@xxxxxxxxxxxxxx > 主题: Re: why objects are still in .rgw.buckets after > deleted > > > > I recall hearing that RGW GC waits 2 hours before garbage collecting > deleted chunks. > > Take a look at https://ceph.com/docs/master/radosgw/config-ref/, the > rgw gc > * settings. rgw gc obj min wait is 2 hours. > > > > > Craig Lewis > Senior Systems Engineer > Office +1.714.602.1309 > Email clewis@xxxxxxxxxxxxxxxxxx > > Central Desktop. Work together in ways you never thought possible. > Connect with us Website | Twitter | Facebook | LinkedIn | Blog > > On 3/16/14 23:20 , ljm李嘉敏 wrote: > > Hi all, > > > > I have a question about the pool .rgw.buckets, when I upload a > file(has been stripped because it is bigger than 4M) through swift > API, it is stored in .rgw.buckets, > > if I upload it again, why the objects in .rgw.buckets are not > override? It is stored again and have different name. and when I > delete the file, all of the objects in .rgw.buckets > > aren’t delete even though I execute radosgw-admin gc process. > > > > I also want to know something about the pool created for object > gateway, why are they created and which role they will play? If anyone > know about these, > > please give me a guide, thanks. > > > > > > Thanks & Regards > > Li JiaMin > > > > System Cloud Platform > > 3#4F108 > > > > > > > _______________________________________________ > > ceph-users mailing list > > ceph-users@xxxxxxxxxxxxxx > > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > > > > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |