I provision storage optimized instances to get the OS on an EBS volume and get a nvme volume for DRBD.
Except the problems appear at reboots. Nitro is randomly switching the order of the disks so nvme0n1 becomes nvme1n1 and the other way around. This blows up the DRBD config unless there is another way to deal with this problem in the config.
Hopefuly someone else bumped into this already.
Blind man speaking of color here, but shouldn’t you be able to just use another identifier to identify your disks?
Have you tried /dev/disk/by-id/XXXXX ? Or by-uuid, by-partuuid?
These days I tend to just default to using LVM and logical volumes, even if I intend to consume the entire backing disk for DRBD with a single logical volume.
Why? It keeps the DRBD resource configurations a little cleaner, no need to define each unique device for each host’s backing disk. Plus, the added flexibility of LVM is there if or when it’s needed.