If I read the release notes for rhel5.3 right, then there's now an option to have lvm mirrors in a clvm environment. In a quest to have a horizontally scalable server pool where each machine both provide its local disks to the pool and mounts the global filesystem provided by the pool, I was thinking along these lines the other day: * each machine has a small private root fs and exports the rest of the disks via the storage block protocol of choice (iscsi/aoe/fcoe) * each machine mounts exported block devices from all other machines * while we can't do raid5 in lvm, we should be satisfied with mirroring; clvm takes care of mirroring between pairs of these network block devices * clvm also takes care of joining together these mirrors in one volume group * clvm also takes care of logical volumes, carved out of this volume group * gfs2 takes care of sharing these logical volumes on a fs level between servers This way a failure of a single server should not affect the working of the whole group, while there'll always be a room for growth both on the cpu/ram side and on the storage/iops side. Opinions? Probably in real world such setup would still have too many bottlenecks to be useful and too many points of failure to be reliable ... But as I see many uses of it, I'd like to see something like that work reliably & fast one day. -- Jure Pečar http://jure.pecar.org/ -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
