Hello. May be because of current implementation of rotten bits detection - one hash for whole file. Imagine 40 GB VM image - few parts of the image are modified continuously (VM log files and application data are constantly changing). Those writes making checksum invalid and BitD has to recalculate it endlessly. As the result - checksum of VM image can never be verified.> 15 ???? 2016 ?., ? 9:37, Gandalf Corvotempesta <gandalf.corvotempesta at gmail.com> ???????(?): > > I was looking at docs. > why bit rot protection is disabled by defaults? > with huge files like a qcow image a bit rot could lead to the whole image corrupted and replicated to the whole cluster > > Any drawbacks with bit rot detection to explain the default to off? > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users-- Dmitry Glushenok Jet Infosystems -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160615/42d83f06/attachment.html>
2016-06-15 18:12 GMT+02:00 ??????? ???????? <glush at jet.msk.su>:> Hello. > > May be because of current implementation of rotten bits detection - one hash > for whole file. Imagine 40 GB VM image - few parts of the image are modified > continuously (VM log files and application data are constantly changing). > Those writes making checksum invalid and BitD has to recalculate it > endlessly. As the result - checksum of VM image can never be verified.I think you are right But what about sharding? In this case, the hash should be created for each shard and not the whole file.
Il 15 giu 2016 18:13, "??????? ????????" <glush at jet.msk.su> ha scritto:> > Hello. > > May be because of current implementation of rotten bits detection - onehash for whole file. Imagine 40 GB VM image - few parts of the image are modified continuously (VM log files and application data are constantly changing). Those writes making checksum invalid and BitD has to recalculate it endlessly. As the result - checksum of VM image can never be verified.>And what about enabling bitrot for small files like emails in a maildir? in this case the bitrot feature with 1 hash for the whole file would be ok. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160630/fee38a76/attachment.html>