Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Felix Frank wrote: > On 06/03/2011 12:26 PM, Sebastian wrote: >> Do you think I would see much performance benefit >> from running my O/S (see next para for details) directly on the bare >> metal? > > I have no idea. Disk I/O is generally the bottleneck in virtualization > setups, but DRBD overhead usually has much to do with network latency > and throughput, which in turn are supposed to virtualize quite well. > > So, if you're facing performance issues, you may want to do > some tests, > but I don't see an immediate need to loose the virtualization layer. OK, performance is not our main concern here, so it seems OK for now. >> /dev/sdb2 ~ 1 TB - /dev/drbd1 (actually LVM so >> /dev/mapper/vg0drbd-filer) - used for NAS /dev/sdc1 ~ 1 TB - >> /dev/mapper/tempnas-tempnasv LV > > Is DRBD on top of LVM or vice versa? > If DRBD is on top, you should have an easy ride indeed. No, LVM is on top of DRBD (although there is only 1 LV currently configured). >> What would be the benefits and >> disadvantages of splitting 1 mounted LV filesystem over 2 DRBD >> devices? > > You mean, create two DRBDs, make both LVM PVs and have an LV > span both? Yes. > Don't. Just don't. Imagine an FS that depends on two > resources that can > each indepentently be available on only one of your HA nodes. > Don't put > yourself into this position. Good point. Back to your plan then! >> When you say "scrub", I presume that is >> optional? > > It is. > > HTH, > Felix Thanks, I will get started on putting DRBD on an LVM then... Kind regards, Sebastian