I currently run Solaris 11.1 in ESXi 5.5, with a HBA (disk controller card) passed though to the OS via VMDirectPath I/O and a raidz pool (called Vault). The OS is running on a single dedicated SSD datastore. This means that if the OS fails I dont have a mirrored version. I dont really want to buy another SSD just to mirror the OS disk. I have looked at Veeam backup solutions however because of the VMDirectPath I/O it will not allow me to back it up while its online (snapshots need to be offline).
My question is, in case of an OS disk failure (the raidz pool is un-affected, seperate disks and controller) would it be better to rebuild the OS from scratch and do a forced import of the pool e.g. 'zfs -f import vault'. OR would it be OK to restore an OS backup that is a week or so old to a new ESXi datastore and I assume this would then carry on as usual. However the data in the pool could have changed, I'm assuming there is nothing on the OS disk that tracks this and would detect the pool as being corrupt/faulted?
Does anyone have a better suggestion for backing up the OS disk, or the best / safest strategy to recover the OS disk and regain use of the zfs pool?