I currently have two datacenters (active/passive) using NFS storage. Backups are done with nightly rsyncs. I want to replace this with RadosGW and RGW geo-replication. I plan to roll out production after Emperor comes out. I'm trying to figure out how to import my existing data. The data already exists in both locations, and it's large enough that I don't want to re-replicate everything. So I have 3 ideas of how to populate and replicate the clusters. #1: Build a cluster in both datacenters. Setup all of the regions and zones, but do not start replication. Execute the same NFS to RGW import in both locations. Start up replication. The documentation (http://ceph.com/docs/wip-doc-radosgw/radosgw/federated-config/) doesn't say that I can't do this. But I don't know if it accomplish anything. Without knowing the internals of replication, this could be a massive speed up, or just useless. #2: Setup both Ceph clusters in the primary data center. Create all regions and zones as they will eventually be, and start geo-replication. Import all data into the primary, letting it replicate at LAN speeds. When the import is done, shutdown the "remote" cluster, and ship all the "remote" servers to the remote datacenter. Turn on the remote cluster and let replication catch up. How long can replication be down and still catch up? What happens when this window is exceeded? This is something I need to test anyway, but I wasn't planning to go quite so extreme in my testing. #3: Setup replication before import, and re-replicate all the data over the internet. Just guessing, I'd say the bandwidth bill will cost about the same as all the time I'll spend testing and executing #1 or #2. If that's true, #3 looks easier, faster, and less stressful. Are there any docs about the mechanism of replication? Knowing how it replication works would answer most of my questions. If it's not documented, it would be worth my time to read the source and write the doc. Are there some other options I haven't thought of? Thanks for any thought or suggestions. _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |