Sorry, but i think this make no sense. I know MDADM is used it for many years and it is not really enterprise. Qnap use it and... oh my god, really problems enough. MDADM sync slow, it must always sync also zero space... a lot of things, that what i think, make no sense to implement that in pve. So the better way is buy memory or reducing the useable memory for ZFS.
I use zfs @home too for three Servers works here fine and it was not so expensive. And if an SOHO are not able to buy that memory then i think there is really a money problem on an other side
But... you feel free to use mdadm with proxmox on traditional shell. I've setuped also some servers with mdadm and it is working fine with low memory too.
I use zfs @home too for three Servers works here fine and it was not so expensive. And if an SOHO are not able to buy that memory then i think there is really a money problem on an other side
But... you feel free to use mdadm with proxmox on traditional shell. I've setuped also some servers with mdadm and it is working fine with low memory too.
@fireon , without any offense, but think for a moment that not all the pmx users live in country where the money is not a big problem. In some countrys you need to work 12 h/day(7 day per week) for 10 euro or less. So imagine for 1 second what will be "not so expensive" for such a guy. And belive me, that I do not see this from some SF movies, but from reality.
Sorry if in my comment I have overreacting.
the problem with mdraid is that you can get easily into a situation where a guest can render your raid faulty (the write semantics of md raid and o_direct is not so nice)
with no way to know which disk is the 'good' one, so mdraid is not supported by us and we will not build it in (for the foreseeable future)
with no way to know which disk is the 'good' one, so mdraid is not supported by us and we will not build it in (for the foreseeable future)