As I understand, linux raid tools operate on partitions only, not on full disks, as hardware based solutions. You can only mirror partition X on partition Y. They can be on any order on the other disk, they can be of any size (the smaller will dictate max size) and they may not be all mirrored if desired. The closest you will get to a hardware based mirror is: 1) have two similar disks 2) copy the partition table of the first to the second, making it have similar partitioning scheme 3) mirror each partition on each other respective partition, in the same order In the end, you will have a disk that is almost perfectly equal, and is even interchangeable with the first (if partitions were in different order, they would not be). Dot not forget to always copy MBR to the second each time you write it on the first. I've seen a product that is a modified debian distro, that installs everything to the first HD and then, on the first boot, "reconstructs" the raid, copying everything to the second one, and rebooting again. I asked why and they said that was the easier way since debian would not support raid on the installer. I think you can do that easily with redhat/fedora, right inside kickstart, but it would still be based on the mirrored partitions logic. In your case, it seems the raid logic would be inside virtual machines, is this correct? You would have double network traffic. You can create one big partition of PV type, and create more small partitions of PV type and make them join the VG each time you grow your disks. The raid would be on each filesystem of LVM. Still a pain in the arse. Hard to switch to raid on the AoE server? andre On 11/21/06, Tracy R Reed <treed@xxxxxxxxxxxxxxx> wrote:
I am setting up a Xen/AoE cluster. Completely diskless cpu nodes which get their root from disk nodes via AoE (an ethernet based SAN) and then fire up Xen domains which also get their disk from AoE. I have all of this working so far with no RAID involved. Now I want to mirror the disk used by the domain over two separate disk nodes. When the kickstart fires up it sees that it has hda and hdb. I want to mirror hda and hdb against each other without creating partitions and then I will create logical volumes inside of the mirror device. But kickstart expects me to create partitions and give them names to be used when setting up the RAID. Any way around this? The reason for doing this is because the machines which export the disk via AoE to the cpu nodes are really just exporting their own local logical volumes which appear as block devices to the xen domain thanks to AoE. I can lvextend the size of the lv on the machine exporting the block device which will become hda to the xen domain and the other machine exporting the block device which will become hdb to the xen domain. Then I do an aoe-revalidate and instantly the xen domain sees a bigger physical volume in which I can expand my logical volumes. If I create one big partition and mirror that I will have a problem because I cannot change the size of a disk partition on the fly when I change the size of the block device under the partition. Any ideas? -- Tracy R Reed http://ultraviolet.org A: Because we read from top to bottom, left to right Q: Why should I start my reply below the quoted text _______________________________________________ Kickstart-list mailing list Kickstart-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/kickstart-list
-- Andre Ruiz <andre.ruiz@xxxxxxxxx> Curitiba, PR, Brasil
