When my laptop disk started to show read errors I searched for a tool capable to save the contents of the disk (multiple systems on it with different file system types). I couldn't find anything which would do a copy of the disk skipping unused areas and tolerating I/O-errors, so I ended up in re-installing all systems after I got a new disk.
The ideal solution would be some clever combination of dd_rescue and tar which could be put on a live-CD like gparted and be able to save the disk in question. To save a 500GB disk with 200GB data on it the tool should require only some 200GB of empty disk space, not the full 500GB of the physical disk to be saved.
over 4 years
Has no hacker:
A normal Xen server today has many single points of failure. Those can be reduced by disaggregating Dom0 into multiple domains (xenstore, driver domains, admin domain). Following this path to it's logical end will lead to a setup where the running domUs won't be depending on Dom0 any more, so Dom0 itself can be made restartable.
This project aims at beginning with a prototype by setting up a configuration with a xenstore domain and a driver domain for networking. This will allow running a pv-domU with network resources only (nfs or iscsi root file system) no longer needing Dom0 to run.
PVH domains are a new guest type supported by Xen being as lightweight as possible (e.g. no emulation of legacy devices via qemu) while taking advantage of the hardware virtualization features of the x86 processor.
As there is no BIOS for a PVH domain booting is a little bit different than for pure hardware virtualized guests. To be able to start such a guest a little shim is needed to gather some information about the environment (especially memory layout) before the standard kernel boot path can be entered. By using a boot loader like grub2 this shim can be avoided as the memory information is already known by grub2 and stored into the so called zeropage according to the multiboot protocol.