Hello all,

I would just like to know what people use as an image based solution to backing up Linux. Not really specifying a distro, just would like to know what people use / philosophies.

Thanks for the helpful replies!

If I really have to make a complete image I use dd in conjunction with a compressor, like bzip2:
1. Write zeroes to the unused parts of a disk, using dd to write zeroes to a file until the disk is full.
2. Use dd to read the partition, pipe the output to the compressor and redirect it to a file on the backup medium.

Since any distro comes with dd and i can choose the compressor this is the most distro agnostic approach, I would think.
Of course there are other approaches with more functionality, like Clonezilla, but I prefer the simple things.

1 members found this post helpful.

Never been one for image backups. I prefer to take a set of tar archives, one per filesystem, that I can use to restore the complete system state, or just individual files/filesystems as necessary. Combine those with the saved output of "sfdisk -d", "vgcfgbackup", and "cryptsetup luksHeaderBackup" and it should be possible to restore everything to its current state from scratch, using only the tools on my slackware installation media.

I can't say I have any good choice. I long for what I perceive as the good old days of ghost.

In the real world we only have a few ideas. One is a bit by bit copy that I use a lot. It has many faults and tips and tricks with it but tried and true.

Second is the file by file method. Many tools can copy off files and a few apps and live cd's work well with this method but again, not me(fool)proof

I work on so many systems and OS's that I use what I can in each place, even if it isn't the best I use what works.

I have tried with clonezilla. But, now-a-days, I do not image based backups anymore. I just rsync to my external HDs. perhaps, it helps that I run no applications that I cannot reinstall. I do backup configuration files as well.

You could try www.mondorescue.org; does a warm backup as opposed to Clonezilla which I believe requires a shutdown as it does a cold backup.
Depends on your requirements.
Databases need special handling if not done cold.

Many (most) prod systems can not be shutdown, so a combination of Having the OS install disks, then something like eg Mondo or NetBackup https://en.wikipedia.org/wiki/NetBackup, with DB hotbackup techniques is usual.
Of course if you have hot swap mirror disks, that works too

Thank you all for the replies!

Im curious about the DD along with bzip2, how would one go about restoring?

The reason why im asking, is that in a complete disaster scenario, I would like to restore the system "at a point in time" and let it resotre an image, than take the time to copy the files, because in a disaster, there are probably other things I need to do while its restoring.

Ill also take a look at mondo rescue, seems like a flexible solution as well.

Thank you all, I appreciate the insight!

What works for me

http://www.linuxquestions.org/questi...s-work-895225/

Take the image (asuming that the partition in question is /dev/sda1 and the backu medium is /dev/sdb1):
1. Boot from a live medium.
2. Mount the partition to image:3. Fill the unused parts of the partition with zeroes, so that the compressed image is smaller:Depending on size of the partition and speed of the disk this can take quite some time.
4. Unmount the partition to image and mount the backup medium: 5. Take the image and compress it:6. Unmount the backup medium and you are done:Restore the image:
1. Boot from a live medium.
2. Mount the backup medium:3. Read the image, uncompress it and write it to the partition:4. Unmount the backup medium (see step 6 above) and you are done.

You may want to also make a copy of the output of so that you can have information on the exact partition size if needed.

Note that I use a blocksize of 16MB for the dd operations, because I found that blocksizes in the range between 8MB and 32MB (depending on your hardware) give me in general a better performance than the default value.

1 members found this post helpful.

@TobiSGD

wow, thank you very much for the detailed instructions. Ill test this right away!

I really appreciate it!

So I tried out @TobiSGD method for using DD and i was able to backup and restore a whole CentOS 6 machine with one image file, very awesome!

There are a few things I wanted to clarify with you though:

The way I did the test, i didnt have to boot into a live media in order to do the backup, i just ran:

and it did the trick, even without writing zeroes, it only backed up the actual usage of the partitions.

also, i didnt specify the partitions of the disk, I just took the whole drive (sda instead of sda1)

When restoring, i ran your other command back to an un-partitioned hard drive with an Ubuntu 12.04 Desktop live CD and it worked, it restored the partitions as well.

again only specifying the hard drive, not a specific partion (sda instead of sda1)

does this make sense?

Thank you everyone for your input!

Zeroing out the free space first allows for a smaller compressed backup. Also consider using ddrescue which might be faster. It will retry if it reaches some bad blocks. It also varies the block size depending on the health of the drive. Bad blocks are zeroed out in the image. Using dd without options will simply fail.

Do you mean you have made an image of the running system? This is not reliable, since the file-systems can change during the backup, which will leave you with an inconsistent backup. Backups that are possibly inconsistent are worthless.

I disagree. Of course you can use ddrescue, but if you have to use it because of a faulty disk it is already to late for a backup, the data is already potentially damaged. If there are already bad blocks on the disk you are not making a reliable backup anymore, you are creating an image for rescue purposes.

Thanks: I was wondering, why the zeros?
That's not really a valid comparison. If you have a tar backup, the backup itself is one file, and can be restored with just one command. As it works at the file level, not the block device level, it gives you enormous flexibility to restore all or part, and to do so to any location (assuming that the backup was made using ./this/that/the-other relative paths) on any device or any point within the directory tree[s].

dd is a lower-level technical tool for when you need a bit-identical, not data-identical, copy. Which I suppose is what "image" means <Blush>. dd will also do stuff like read or write a given number of bits, convert block sizes, etc.

Tar is your swiss-army knife: dd is your leatherman multi-tool.

The problem with running dd on a live file system is that things can change while the backup is running e.g. changes to log files. Also you will backup the the contents of /sys and /proc which are automatically generated at boot.

Personally, I like dd images of my Windows machines. If you restore from the image after a virus invasion it is comfortable to know that the sucker has been completely eliminated from the disk.