Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 09/28/2017 02:31 PM, Gandalf Corvotempesta wrote: > Hi to all, > Some questions about DRBDv9 [...] > > c) what if I have to aggragate multiple resources to increase spaces ? > Can I use LVM with something like this: > > pvcreate /dev/drbd0 > pvcreate /dev/drbd1 > vgcreate vg0 /dev/drbd0 /dev/drbd1 > lvcreate ....... > > to create a single LV on top of 2 drbd resources ? It's possible, but for many use cases, it is not a good setup. There are a couple of potential problems with this setup: 1. Both DRBDs must be uptodate for the volume group to function 2. LVM must be reconfigured to ignore the backend device and wait for DRBD to start 3. The volume group and all of its volumes are only available on one node at a time, unless you use a Dual Primary setup, which is much more fragile than a Single Primary setup (basically, every loss of the replication link immediately results in a split-brain situation), also you would need CLVM with such a setup, and even then, the volume group would only be usable on two of your three nodes at a time (it's called "Dual Primary", not "Multiple Primary"). Therefore, the sane way for many use cases is: pvcreate /dev/sdc pvcreate /dev/sdd vgcreate vg0 /dev/sdc /dev/sdd lvcreate ... and then replicate each LV with a separate DRBD resource. With such a setup, different volumes can be active (Primary) on different nodes (e.g., it is easy to build an active-active 3-node cluster) and still be replicated to all three nodes, easy to extend in size, etc. The only drawback would be that a loss of one of the physical volumes results in loss of the volume group on that node (but you have two other nodes with the same data, and the DRBD devices are still available on all nodes as long as the replication link is up). Adding only RAID (1, 5, 6, 10, 50, 60) devices as physical volumes to a volume group can obviously also prevent loss of an entire VG on a single node in case of a single device failure. > f) a 3node cluster is still subject to splitbrain ? A cluster with any number of nodes may experience split brain situations whenever cluster communication fails and no fencing mechanism is in place. br, -- Robert Altnoeder +43 1 817 82 92 0 robert.altnoeder at linbit.com LINBIT | Keeping The Digital World Running DRBD - Corosync - Pacemaker f / t / in / g+ DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.