[Linux-cluster] Recovery disaster

Pierre-Alain RIVIERE pariviere at ippon.fr
Thu Jul 10 14:02:39 UTC 2008

Hello everyone,

We're using Xen for about a year in my organization and I want to make
profit from summer to improve our infrastructure. First step : plan a
full recovery disaster procedure. It's not only related to Xen (only a
few actually) so I've allowed myself to post on both Xen and Linux
cluster lists.

My infrastructure is built as followed :
- One software SAN built with Openfiler (http://openfiler.com) : big
disks, RAID 5E, redundancy on power supply, network, cpu and RAM.
- N Xen Dom0 (actually 3)
- The same iSCSI volume is mounted on each Dom0 and we're using CLVM on
it. A PV equals a DomU disk.

It works pretty well and now I would like to rebuilt my SAN as quickly
as possible in case of problem (big hardware failure on the SAN).

Here how all these stuffs work together :
|---------Openfiler------|            |---------Dom0---------|
PV -> VG -> LV -> iSCSI -> network -> PV -> VG -> LV->Xen VDB

PV : physical volume
VG : volume group
LV : logical volume

- We use the LVM layer (DomO side) on top of another LVM layer (SAN
side) and performance are good until now. Do you know some caveats
about this usage? Is there's any reason for me to switch to a network
aware filesystem?
- Can I dd a snapshot of the iSCSI volume on the Openfiler box, send it
to a tape driver and expect a dd back to a identical LV to work?
- Same question if 2 or more LVs on the Openfiler box are aggregated
together with CLVM (and though iSCSI) on the Dom0 side.


More information about the Linux-cluster mailing list