[PVE-User] offline VM migration node1->node2 with local storage
Aaron Lauterer
a.lauterer at proxmox.com
Mon Feb 17 09:37:48 CET 2020
On 2/16/20 11:53 AM, Roland @web.de wrote:
> for my curiousity, in a cluster i can have "rpool" on both nodes (both
> are local to each node, i.e. are different pools but with same name) and
> i can blazingly fast offline migrate VMs from rpool on node1 to rpool on
> node2 (as it's done via zfs snapshot).
>
> but i cannot add a second pool "hddpool" on different disks on both
> nodes with gui, as i get error "storage ID 'hddpool' already defined
> (500)" if i want to add 2nd zfs pool to node2 with the same name.
>
> what i can do is adding pool via commandline on node2 and then remove
> "nodes node1" from strorage.cfg (manual edit) to make it analogous to
> rpool - i can use hddpool like rpool after that.
>
> i wonder if this is the proper "way to go"!?
There is a "Add Storage" checkbox below the name field when you create a
new zpool via the GUI.
After you created the pool on the first node, you have to uncheck it
because the storage configuration is shared among all cluster nodes and
already exists.
>
> anybody running proxmox cluster without shared storage as larger scale?
>
> what i try is developing a concept for running a maintainable vm
> infrastructure without need for HA shared storage&networking (as we have
> vm live migration including disks and on server defect, i simply could
> pull all disks out and put into a spare machine of same type)
Have you taken a look at the built in Replication to have a copy of a VM
on another node?
How many nodes do you want to have in your cluster? To be honest, ZFS
and replication is a nice setup for a smaller 2 node cluster (with
QDevice for quorum). Once you have more nodes it gets complicated quite
quickly. A shared storage does not only give you HA, should you need it,
but simplify the administration of the cluster.
If you want to avoid an expensive storage appliance that supports HA,
have a look at the Ceph integration of Proxmox VE.
>
> regards
> Roland
>
>
More information about the pve-user
mailing list