Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
It took some digging before I realized that the formula for external meta-disk storage is 128MB * DRBD devices. I'd suggest changing the man page from: You can use a single block device to store meta-data of multiple DRBD devices. E.g. use meta-disk /dev/sde6; and meta-disk /dev/sde6; for two different resources. In this case the meta-disk would need to be at least 256 MB in size. To something like: You can use a single block device to store meta-data of multiple DRBD devices. E.g. use meta-disk /dev/sde6; and meta-disk /dev/sde6; for two different resources. The storage required is 128MB * number of DRBD devices. Then where I looked most closely was: http://www.drbd.org/users-guide/ch-internals.html Maybe add a section saying: 18.3 Calculating external shared meta-disk size = 128MB * DRBD devices. Googling around the most authoritative source I found for this was: http://lists.linbit.com/pipermail/drbd-user/2009-July/012443.html Seems like other people had the same problem: http://lists.linbit.com/pipermail/drbd-user/2009-July/012444.html My approach before I knew about the restrictions on the shared meta data (and the normal formula not appling was: drbdnode: 1TB disks A B | C D partition 1 raid-1 | raid-1 partition 2 LVM LVM | LVM LVM So I was hoping the RAID-1 on disk a and b could hold the metadata for any DRBD devices on partition 2 of disks C and D. So naturally the RAID-1 on disks C and D would hold the metadata for any DRBD devices on partition 2 of disks A and B. So if any 1 disk gets busy the metadata writes don't decrease the sequential write performance. With this configuration if I lose any one disk I just have to rebuild 1 TB. Without the RAID-1s the metadata would not be replicated and I'd have to rebuild 1TB because of loss of data, and a 2nd TB because of loss of metadata. Alas, since I want to be able to have an arbitrary number of logical volumes for virtual machines I can't presize the raid-1.