On 10/06/2020 4:38 pm, Mark Adams via pve-user wrote:
The simplest thing to set also is to make sure you are using writeback cache in your vms with ceph. It makes a huge difference in performance.
Chiming in - doing some testing with a 5 node ceph/proxmox cluster here. Basic spinners and 4*1G eth, LACP:tcp-balance.
enabling KRBD on the ceph pool made a huge difference - I presume that uses the rbd kernel driver?
-- Lindsay _______________________________________________ pve-user mailing list email@example.com https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user