Am 21.03.2007 um 23:02 schrieb Sridhar Ramaswamy (srramasw):
Don't we have a potential solution using GNBD and CLVMD?
So you've bunch of PCs with hard disks.
1) Each PC create LVM PhysicalVolume (pvcreate) on its storage device
2) Each PC exports the physical device using GNBD non-cached
3) Each PC imports other PCs storage using GNBD
4) Create a VolumeGroup combining GNBD devices (like
/dev/gnbd/gnbd_pc01) and local physical devices (/dev/hda8). CLVMD
will
take care of distributing LVM info among cluster nodes.
5) Create GFS on top of an logical-volume created off this above
VolumeGroup
Would such a thing work alteast as described above? Such a setup will
aggregate storage off these PCs into a single volume.
Ofcourse, even if it does "work", it is a different question
whether it
make sense to deploy such a solution. Major downside is - even if
one PC
shuts down, the whole VolumeGroup would be offline and GFS won't be
available. Fencing doesn't apply here.
;--)
Usually, "distributed" still implies some sort of redundancy.
Unless you're Google and your data consists of copies of downloaded
webpages, losing some of your data is not an option.
Rainer
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
