I am using latest proxmox, latest kernel (pve-kernel-3.10.0-2-pve because 2.6.32 did not work well with adaptec 6405). When doing backup to NFS storage, machines always start freezing (INFO: task jbd2/vda1-8:299 blocked for more than 120 seconds, ... inside virt. machines). I had the same problem with CIFS share. This is caused by very bad vzdump design - when backup storage is under load, write speed is about 17 MB/s - this is problem for vzdump "out of order in memory backup". Manual backup using LVM snapshots and "dd if=/dev/pve/mujsnap-107-disk-1 bs=100M | lzop -c > /mnt/pve/backup/muj/mujsnap-107-disk-1.lzo" works well. Is there any change to switch back to LVM snapshot in vzdump? Way the vzdump works now just brings me problems, even if nfs storage is not under load and write speed is more than 50 MB/s. So: 17 MB/s, iowait 12% => very big problem for vzdump design, machine freezes BUT NO problem with LVM snapshots, everything is working well
↧
