Use checkpoint feature along with Snapshots at slave side to make sure each days snapshot contains all the changes from that day. When you set checkpoint, you can watch the status of checkpoint completion using geo-rep status. Once checkpoint complete, it is guaranteed that everything created before Checkpoint Time is synced to slave.(Note: It only ensures that all the creates/updates done before checkpoint time but Geo-rep may sync the files which are created/modified after Checkpoint time) Read more about Checkpoint here http://gluster.readthedocs.io/en/latest/Administrator%20Guide/Geo%20Replication/#checkpoint regards Aravinda On Monday 21 November 2016 01:36 PM, Gandalf Corvotempesta wrote:> > anyone? > > > Il 19 nov 2016 10:36, "Gandalf Corvotempesta" > <gandalf.corvotempesta at gmail.com > <mailto:gandalf.corvotempesta at gmail.com>> ha scritto: > > Would be possible to setup a geo-replicated cluster to be used as > backup? > Obviously deleted files on master would be also deleted on the > replicated, so this won't be feasable to be used as backup, BUT if I > schedule a snapshot every night (on the replicated node), I can use > the snapshot as backup. > > I case of restore, I have to mount the snapshot e restore from there. > > Something like this, on the geo-replicated node: > > # gluster snapshot create backup_$(date +%Y%m%d_%H%M) gv0 > > in a cronjob > > Any drawbacks ? > Backing up a multi-terabyte or multi-petabyte would be impossible with > standard rsnyc or any other backup tools. > > > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20161121/e58be61a/attachment.html>
2016-11-21 15:48 GMT+01:00 Aravinda <avishwan at redhat.com>:> When you set checkpoint, you can watch the status of checkpoint completion > using geo-rep status. Once checkpoint complete, it is guaranteed that > everything created before Checkpoint Time is synced to slave.(Note: It only > ensures that all the creates/updates done before checkpoint time but Geo-rep > may sync the files which are created/modified after Checkpoint time) > > Read more about Checkpoint here > http://gluster.readthedocs.io/en/latest/Administrator%20Guide/Geo%20Replication/#checkpointThank you. So, can I assume this is the official (and only) way to properly backup a Gluster storage ? I'm saying "only" way because it would be impossible to backup a multi terabyte storage with any other software.
Is it possible to set the checkpoint on the master node from the slave node where the snapshot will actually be done? Because I can see the problem that you have a "backup" script on the slave node which actually does the snapshot but somehow it has to run/set the checkpoint on the master node. Somehow these two tasks need to be coordinated. First set the checkpoint, check if checkpoint is successfully and then and only then create the snapshot. How do you coordinate these? Best, M -------- Original Message -------- Subject: Re: [Gluster-users] geo replication as backup Local Time: November 21, 2016 3:48 PM UTC Time: November 21, 2016 2:48 PM From: avishwan at redhat.com To: Gandalf Corvotempesta <gandalf.corvotempesta at gmail.com>, gluster-users <Gluster-users at gluster.org> Use checkpoint feature along with Snapshots at slave side to make sure each days snapshot contains all the changes from that day. When you set checkpoint, you can watch the status of checkpoint completion using geo-rep status. Once checkpoint complete, it is guaranteed that everything created before Checkpoint Time is synced to slave.(Note: It only ensures that all the creates/updates done before checkpoint time but Geo-rep may sync the files which are created/modified after Checkpoint time) Read more about Checkpoint here http://gluster.readthedocs.io/en/latest/Administrator%20Guide/Geo%20Replication/#checkpoint regards Aravinda On Monday 21 November 2016 01:36 PM, Gandalf Corvotempesta wrote: anyone? Il 19 nov 2016 10:36, "Gandalf Corvotempesta" <gandalf.corvotempesta at gmail.com> ha scritto: Would be possible to setup a geo-replicated cluster to be used as backup? Obviously deleted files on master would be also deleted on the replicated, so this won't be feasable to be used as backup, BUT if I schedule a snapshot every night (on the replicated node), I can use the snapshot as backup. I case of restore, I have to mount the snapshot e restore from there. Something like this, on the geo-replicated node: # gluster snapshot create backup_$(date +%Y%m%d_%H%M) gv0 in a cronjob Any drawbacks ? Backing up a multi-terabyte or multi-petabyte would be impossible with standard rsnyc or any other backup tools. _______________________________________________ Gluster-users mailing list Gluster-users at gluster.org http://www.gluster.org/mailman/listinfo/gluster-users -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20161122/c0df928b/attachment.html>