Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hi there, I have to plan some 10 Linux Servers in 1 Location. The Servers do webhosting, typo3-hosting, so mysql-stuff, mailserver, antispam, imap mailboxes,virus checks and all that stuff. The average needed diskspace per host is about 25 GB. The need for heartbeat and immediate automagic failover is not there. I want to consolidate these servers to 2-3 xen hosts and normaly I should use a SAN, but a SAN is too expensive for this setup. So I played around with AOE, ISCSI, GNBD and such stuff but they are all not fast enough and not reliable enough for me. So my conclusion is, that ethernet-storage is not an option (except drbd which is fast and reliable). Here drbd comes to my mind. My idea is basically: -Run the stuff on the Xen-Hosts and the VMs under LVM on cheap local harddisks for performance reasons. -Replicate each logical volume via drbd to one or 2 storage hosts so that we have a current-copy in case of a crash. -In case of a crash of a xen host I could copy the logical volumes (xen vms) from the storage host to another host with another cheap harddisk or just mount the drbd partition from the storage host and continue running the server. My next idea is to speed up backups. The Network Amanda Backups of the xen hosts take a lot of time and I don't want to backup Servers via Network anymore. With the above idea i might be able to backup the vms right from the Storage hosts. Maybe I could even run the drbd in master-master mode and backup a snapshot of the drbd partition instead of the drbd partition. What do you think about the above ideas ? Do they make sense ? Thanks Christoph