Hi,
i installed proxmox 1.9 on an hp ml350g6 with a 3ware controller (with BBU) and corsair ssd force 3 gt disks. I made 2 ext4 partitions and all the optimizations for ssd drives. I made my VMs with qed format. I tried to modify my qemu-server's configs, adding "cache=writeback" feature which enables the writeback simulation over the hypervisor and so i made some benchmarks. Obviously, with cache=writeback enabled,VMs are much faster, specially on 4k writes and reads.
Is it safe for a production-ready system this option? I know how it works, but i have a doubt: if a power loss happens, the VM can finish to flush thier write operations? Or the disk image will be corrupted?
Thanks
i installed proxmox 1.9 on an hp ml350g6 with a 3ware controller (with BBU) and corsair ssd force 3 gt disks. I made 2 ext4 partitions and all the optimizations for ssd drives. I made my VMs with qed format. I tried to modify my qemu-server's configs, adding "cache=writeback" feature which enables the writeback simulation over the hypervisor and so i made some benchmarks. Obviously, with cache=writeback enabled,VMs are much faster, specially on 4k writes and reads.
Is it safe for a production-ready system this option? I know how it works, but i have a doubt: if a power loss happens, the VM can finish to flush thier write operations? Or the disk image will be corrupted?
Thanks