Importance of validating backup images

First of all, a quick reminder that the validation process is only about establishing the integrity of the backup, not the fidelity.

It is important to validate because of a couple of scenarios:
- when you do a chain of backups, some user errors, hardware issues or ATI software issues (eg case of deleting some backups) can compromise the chain of backups,
- a single file can get compromised by otherwise benign operations (like a windows file copy).

A failed validation is often the indication of a problem, but not systemically that the image cannot restore, or cannot be used.
Vice versa, it is not because an image validates that it can be restored (because validation is only about the integrity, and many other things can hinder restoration).

Successful validations are like insurance: they provide peace of mind. They also increase your level of confidence that you can restore.

Some advanced users don't validate because they have enough backups that even if one gets compromised, they have always another one to go back to.

Some other users don't validate but restore their backups on a regular basis while rotating the system disks. This is kind of the ultimate peace of mind since they use an actual restore to validate the backup.

I set up a validation schedule so that each full backup is validated at least once. From time to time, I boot on the recovery CD and restore a couple of files from my last full backup.

I always validated. I had a bad SATA cable and it was caught by the validation process.

Sectors can go bad for various reasons but if there is a problem with marginal magnetization on a particular sector it isn't likely to be caught with an immediate validation. I did have some archives fail later on a notebook but this was due to bad sectors probably caused by a head crash. Pat L's implication is correct, you should always have more than one backup in case you have a problem with the current one. In my notebook episode I had to go back 2 or 3 archives before I found a good one.

PCs do very little if any error-checking on writes to disk or RAM and nothing approaching the rigorous checking the validation routine does. TI writes 4000 checksums/GB of data and everyone has to compare perfectly or the archive is declared corrupt. This also checks the data path right to the processor and RAM. PCs typically pickup data errors on reading or in the case of RAM when the machine goes haywire - there is no check on data read from RAM unless you are running a high-end machine with parity or ECC memory.

If you are running with zero validation errors then you probably don't need to validate every archive but it's easy to do but just takes a bit of time.

You really need to ensure the TI recovery CD will validate your archives if you haven't done so. The recovery environment is Linux and sometimes the drivers are not the best fit for your hardware. If the CD can't validate it means it can't read the archive properly and an image restore will be impossible. Once you know this works then you don't have to keep checking it unless you make a hardware change - for example a different external drive or add-in disk controller card etc.

Seekforever wrote:

PCs do very little if any error-checking on writes to disk or RAM and nothing approaching the rigorous checking the validation routine does.

That's because writes to disk or memory are extremely reliable. The chance of having your primary drive fail AND a problem with the most recent backup simultaneously is very, very low. I haven't even had a hard drive fail in 20 years of heavy computing.

Still, when something does eventually go wrong, it's not going to be pleasant. That's why I finally started making backups a few years ago by cloning partitions directly to an identical drive. I only started making images in the last week, to save space with compression and take advantage of incremental backups.

The two validations I ran so far were clean. I'm just not sure about the value of running an incremental backup for 15 minutes and then validating for two hours, even seamlessly in the background. I'll probably keep validating until I'm more comfortable and then do it less frequently or stop completely. I'll also ensure I can validate from the recovery CD as suggested. Actually I have both True Image and Disk Director set up in the Windows and GRUB boot loaders to load from the hard drive with a bootable flash drive as backup, but that's another story.

Thanks for the input (both of you).

I've had RAM fail on a WHS and it corrupted the backups. The images being validated were caught, but the others were lost. I usually validate my backups because I want to know for sure that they are good at the time of the backup. I don't want to find out when I need it that it was bad at creation.

You could run a validation once a week or once every two weeks. That would probably be often enough if everything seems to be running smoothly. My main point is that hardware can fail without notice or warning. I've lost a number of drives over the years, including several durring a backup. From my recent experiences, the quality of drives have decreased greatly in recent years (they're a lot cheaper, though).

If you have validated the backup at least once, you'll know that the Full image is good. The Incrementals may be less important and validations can be run as needed. Depending on your schedule, you may not even bother. For example, if another Full is being created in a week or two weeks.