LVM Frustrations in AWS
Our security folks require that any EL6 systems we deploy have a specific minimum partitioning scheme for the root filesystems. The most practical way to meet these requirements is to use LVM. Under a 'cattle' model, using LVM for root is a non-issue ("botched a config file in /etc and broke the system? No problem: nuke it and re-launch"). Unfortunately, many of the applications we support haven't reached the point of re-launch being quicker than a repair-attempt).
Further complicating things is that our security rules mean that application owners only have a very limited set of AMIs they're allowed to launch (i.e., can't just launch a community/MarketPlace AMI that doesn't use LVM to act as a fix-host). If they want to move their root disk to a fix-host, there are LVM collisions because the LVM objects' names and UUIDs.
The name issue is fairly trivial to solve with about four commands and a reboot. The UUID problem, however, is a blocker. Even with non-colliding LVM object names, instances that launch from the same AMI have the same UUIDs. This confuses the hell out of the LVM tools making it so that doing a pvchange --uuid /dev/path against the (to-be-imported) disk fails.
Anyone know a good way to alter the PV UUIDs outside the scope of the LVM tools, or is this just a total horror-show to try to sort out?
