Greetings all, I have a question about scalability and redundancy and am looking for some pointers/advice.
Suppose I have setup a 4-node distributed-replicated gluster configuration that was created using the command, "gluster volume create test-volume replica 2 transport tcp server1:/brick1 server2:/brick1 server3:/brick1 server4:/brick1" >From what I understand after reading the gluster docs and doing my own lab >tests, it seems all the data contained on the replica pair will be lost of >both servers in the replica fail. For example, if I loose nodes 3 and 4 at >the same time, all the data contained on those nodes is lost. How can I get >gluster to identify a node outage and replicate the data from server3 to >servers1&2 if server 4 fails? Is there some sort of command I can run so the >data on node 4 gets replicated to node 1 or 2? I have read some threads about creating a 3-node replica, but that only gives me 33% of the disk space since we are replicating the data to two other nodes. In addition, if I need more space, I have to add 3 more nodes to the pool since the replica count is 3. Seems like a big waste of money. My goal is to build a storage infrastructure using Gluster to hold virtual machine images and other data for our infrastructure that can survive multiple node failures. These servers are high-quality SuperMicro servers with LSI RAID controllers, plenty of RAM and CPU. However, in the very unlikely event we loose 2 servers, I don’t want to loose any data. Thanks, -Ron _______________________________________________ Gluster-users mailing list [email protected] http://supercolony.gluster.org/mailman/listinfo/gluster-users
