On 29/10/14 17:40, Dr. Boris Neubert wrote: > > > Am 28. Oktober 2014 21:46:26 schrieb Adam Goryachev > <mailingli...@websitemanagers.com.au>: > >> On 29/10/14 07:00, xpac wrote: >>> Currently I have BackupPC set to simply backup the entire directory which >> contains the .vmsd, .vmx, .vmdk, etc files. >>> Is there a "better"/recommended/best practices way to do this that is >> different than what I am currently doing? >> You could in theory take a snapshot of the LV that the disk image sits >> on, mount the snapshot, then create the "chunks", and then umount/delete >> the snapshot, or any similar method, though a restore will produce an >> image of a "crashed" machine (like power failure crashed). > That's what I do except for creating the chunks. The drawback about the > large files is that the initial backup takes days to complete. In addition, > I use the snapshot feature to be able to revert to previous version of the > VM. I therefore keep only one full backup at all.
Actually, with large files and small (relatively small compared to the size of the file) changes is that even an incremental backup can take a very long time (again compared to the size of the changes). I try not to keep snapshots, because they affect the performance of the VM so significantly. backuppc pre backup sets up the snapshot, mounts the VM disk, etc, then the backup is completed to backup the actual files of the VM, and finally the post backup umounts and deletes the snapshot. Back to the images though, each day (not controlled by backuppc, just scheduled), the image split/etc is done. I just use the unix split, and also a small C program I wrote which simply slows down the IO (ie, I set the number of bytes/sec of STDIN that is sent to STDOUT so that I don't utilise all available disk IO and slow down all the running VM's), and write to another directory. So yes, it uses up double the space of disk capacity (but capacity is cheap, performance is expensive). Therefore when backuppc runs, it will always see one set of the split files, and these are mostly unchanged (even though the timestamp changes), that just means rsync needs to read the full file on the backup client, backuppc doesn't need to with checksum-cache, but since there is no content changed, it is very quick. I can only suggest you test this yourself to see the difference it can make on your systems, for me it was significant. In addition, I also copy those small split files to another remote system for DR purposes, where there are multiple directories of the small files. In a DR situation, I just need to cat all those files back together, and then boot up the VM. > I plan to refine this by additionally use BackupPc on certain virtual disks > with user data. I don't know what you mean by "use *backuppc on virtual disks* which also contain user data" or if you mean "use backuppc to *backup virtual disks with user data*".... Hopefully the second :) Regards, Adam -- Adam Goryachev Website Managers www.websitemanagers.com.au ------------------------------------------------------------------------------ _______________________________________________ BackupPC-users mailing list BackupPC-users@lists.sourceforge.net List: https://lists.sourceforge.net/lists/listinfo/backuppc-users Wiki: http://backuppc.wiki.sourceforge.net Project: http://backuppc.sourceforge.net/