Hi, I'm still a little confused as to what the best option currently is to have a simple setup with 2 glusterfs servers which would achieve both redundancy and load-balancing. The "complex" solution seems to create two main volumes plus one namespace volume on each server, then use afr to have 3 RAID-1 type volumes, then unify the two main ones, using the other as the namespace. This seems overkill, and means that all replication "goes through" the clients. This also only provides redundancy (failover) if the clients are pointed to a hostname resolving to both server's IP addresses. It also "splits" the files on each server to achieve round-robin, which again seems overkill to me. The "simple" solution is to create one main volume on each server, then use some external tool to sync files between both servers (rsync, csync2...), then use the same hostname to multiple IP addresses trick. The biggest downside here is that files aren't immediately available, since they're created on only one of the servers. My ideal solution would have : - Clients access the servers using rr or any other algorithm. - Clients stop accessing a given server if they can't use it anymore. - Servers replicate their data directly between each other. Am I missing any better solution in 1.3.x? Matthias -- Clean custom Red Hat Linux rpm packages : http://freshrpms.net/ Fedora release 8 (Werewolf) - Linux kernel 2.6.23.9-85.fc8 Load : 0.48 0.53 0.50
