Hello, I'm running Proxmox PVE with GlusterFS storage over Infiniband and have noticed that the way proxmox assigns drives to virtual machines over glusterfs (-drive file=gluster://..) causes qemu and glusterfs to communicate over TCP. This is not only slower but also more CPU intensive for RDMA enabled interconnects.
I was thinking of (actually implemented it on my servers) adding a "transport" option to storage.cfg. In my implementation transport is optional and if not specified defaults to the current behaviour. Possible values are tcp, rdma or unix (haven't tested the latter though). So my question is are you (Proxmox) guys interested in implementing this feature? I gotta clean up my patch a bit, as currently it's not against git version. Regards, Stoyan _______________________________________________ pve-devel mailing list pve-devel@pve.proxmox.com http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
