Should I prefer some specific sort of backup tool in a dying hard drive situation?
 

I currently have an HDD that's disabled/turned off because it was receiving a SMART warning "BAD, backup and replace".

I was wondering if some backup strategies are better/safer than others in this sort of situation. Such as simply mounting the HDD, searching for the most important stuff and copying it just as if it was a normal drive, versus using something like dd or fsarchiver.

I thought that dd-type things would be superior as they would (I guess) access the disk in a more crude manner that would perhaps require less mechanical action than searching and copying files as a mounted filesystem, even though it would copy everything rather than just a selection. And by requiring less mechanical action I guess it reduces the odds of the disk actually dying during backup.

In the other hand I've read in some forum, maybe even here, someone recommending to do a backup (I guess it was in a similar situation, failing disk) with the mounted filesystem, but I don't recall the reasoning behind it, nor could I find anything relevant yet.

A semi-unrelated question is whether copying stuff on a mounted file system can be done in significantly different ways, with different software. I've once used grsync for some copying of this sort and I had the impression that it was copying way faster (the same disk, partition type, and mounting options) than with a file manager (the one I use more regularly is Konqueror). But it could well be just an "illusion" from the different progress displays.

If it's a pc & not server then I would get a new hdd & install new OS then mount & copy important docs & whatever I thought was important to me, then possible checking for other alternatives of backing up or trying to preserve existing OS. Just my personal opinion.

It depends on the type of failure and the amount of data you have to recover. If you have to recover only 500MB of data on a 2TB disk the file-based approach may put much less stress on the disk (in the case that bad blocks aren't preventing this approach). With larger amounts of data and the possibility of bad blocks preventing a file-based copy I would recommend to use specialized tools, like ddrescue, dd_rescue or other forensic tools, for the recovery of your data.

You may have noticed that I didn't use the word backup in the above paragraph, just because what you are doing is not a backup. A backup is done on known good data, not on failing disks with possibly corrupted data. The best backup strategy is making backups while your hardware works, for the case that in the future it possibly doesn't work anymore.
Making "backups" when the hardware is already failing is like fastening the seatbelt after your car crashed into a wall.

Thanks, EDDY1 and TobiSGD. I was thinking about doing something like that, try copy "normally" whatever I find the most interesting to keep, and then try another approach for a more careful analysis of "everything". Regardless of the "best strategy" regarding the odds of ultimate hdd failure, it's perhaps safer as I've never used anything like dd or other stuff, so I could simply fail at my first attempt.


Yeah, I'm seriously considering buying 2 new hdds instead of just one right away, and leave one always nearly empty, just swap and readily stuff I don't really care much. Funny that after 20 years of having PCs with hdds, only in the last few years I've had two hard disk failures/warnings, both from new hdds. It used to be that the hdds would outlive the PCs, and I'd just replace them when they were just too small for current standards. I guess my power supply and/or energy grid of my current location are probably the culprit(s).