I know that with Ploop you lose disk space as files are added/deleted within a container and you have to manually compact the container, but i have also read that if you reboot or there is a crash the containers can become nonrecoverable.
Are there ways to properly reboot OpenVZ servers to avoid corrupted containers?
Even with an unclean shutdown of plooped containers, it's usually enough to do a fsck. It isn't much different from shutting down real hardware during write operations. Since I'm guessing you'll be using a filesystem with a transaction log on those ploops, you shouldn't run into more issues.
As Brian mentioned, a proper shutdown of the host node will leave your containers clean and their ploops unmounted.
If you're using directory-based containers that use the same filesystem as the host node, you already have a small chance of corruption now. I think in practice there isn't much difference, only that you might have to consider recovery time. Many ploops might take longer to fsck than one host node's filesystem and might require a lot of manual interaction.
On the other hand, if your host node has a very large file system, fscking that could take a very long time, and that leaves all those containers down during fsck. Containers using ploop could be started in a staggered fashion after the host node comes back up, provided the host node itself has nothing to fsck.
You could mitigate that by having some sort of high-availability setup where the host nodes load their container data from central storage and switch each other on/off in cases of failures, but I guess that's going too far.
What are your reasons for considering ploop? We've considered it for faster performance through NFS, where the many small files in a container really slow things down. But since you're mentioning filesystem corruption, that's probably not your scenario.
No, you have to wait until each ploops are stopped properly.
ploop containers must be started to access their files through the root directory of theirs.
simpfs containers are directory structures similar to chroot so it can be accessed regardless on/off.
I manage thousands of containers and I can't manage to waste time waiting. I must wait for ploop containers to be up whenever I need to access or do the migration, or wait until it is off, to manage host that has ploop containers. also, if ploop gets damaged then you have to start all over, it's gone. as I experience this damages not once, or twice, I call it an unreliable system regardless of how good it should be.
reboot) will have OpenVZ stop all containers then restart them again on boot. An ungraceful reboot (crash,reboot --force) means a chance of corrupted containers.