Hi, I am new to Ceph and just trying to get to grips with all the different concepts. What I would like to achieve is the following: 1. We have two sites, a main and a backup site. The main site is used actively for production, and the backup site is there for disaster recovery but is also used for debugging issues. In particular, the files of the production site are available on the backup site as well and can be accessed. 2. We would like to have near-real time replication from main to backup site of a big file system. The files on the backup site should remain read-only. In case of a disaster, the backup files should become writable but this is not a strict requirement. The network connection from main to backup is a slow(er) WAN connection. 3. A replication level of 2 (all files stored twice) is sufficient for the main environment and a replication level of 1 (only one copy) would even be sufficient for the backup (although 2 would also be good). 4. Replication on the main site is done by two machines where each machine has all the files. When one of the machines goes down, the other one can still continue to read/write the data 5. FIles are stored in clusters where different clusters are logically independent groups of files. Now, looking at the concepts of Cepth I am wondering what would be a better choice. Using just the block device would be easier to setup in Ceph I guess, but then access to the files by different machines on the main and backup environment would be more complex since it would require a cluster filesystem suse as GFS or OCFS2. So therefore the setup would probably resort to the Ceph Filesystem. Looking at the mirroring from main to backup site this would require the RBD mirroring functionality I suppose but this replicates at the block device level. Would it still be possible to access the filesystem on the backup environment in this way? Looking at the main setup, I would use 2 OSDs per cluster and 3 monitoring processes to deal with one a failure or one machine right? One solution I am thinking of is to have one monitoring process per OSD (co-located with the OSD perhaps) and then one additional one in the network. What are your thoughts about this? I am assuming this is a relatively standard set of requirement with file system redundancy on a single site and replication of files over a WAN from a main to a backup site. Are there any standard recipes/blueprints that I could follow to set this up? Cheers Erik -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.ceph.com/pipermail/ceph-users-ceph.com/attachments/20170306/2c39caa5/attachment.htm>
 |