@garrett@st.rawddrescue is designed to re-try bad sectors multiple times, thus trying to rescue as much of the available data as possible. From my experience, using ddrescue once to clone the failing disk is safer then trying to copy files directly, as ddrescue optimizes the disk accesses to the strictly minimum required to copy the data, and hence limits the possible additional damages.
Ok, that’s cool. I wasn’t aware of its optimization on disk accesses when using it before. Thanks for sharing!
I have had drives that slowly died (getting a few bad clusters over usage, sometimes immediately detected, sometimes not so much — just with IO errors in logs) and those that worked for a while and then just stopped. It’s the latter type that someone needs to be careful around and optimize which data to copy ASAP. But you really have no way of knowing what your drive will do.
I did have a disk around 15 years ago that fell in this category of a sudden cliff of brokenness. I had most things backed up, but I did have a few new things not backed up. (Whoops!) Anyway, I managed to pull down a Documents/ directory and prioritized grabbing directories of my photos, in reverse chronological order. Then I started to grab some more data that I wasn’t sure was backed up and the drive died and stopped responding. I let it sit for a while, plugged it back in, grabbed a little bit more off of it, then it completely died, never to be used again.
Since then, most of my hard drives have died slow deaths, so the image-first would’ve worked there. (I’ve been more careful to have up-to-date backups anyway.) I have had two instantly die without any warning in the past few years though — both USB. One because I bumped it a little while it was in use. (Whoops!)
Yeah, I must always choose places to live around some magnets in the ground, have a literal magnetic personality, have found a 2-leaf clover, attract a lot of space particles, or something else to have so many hard drives die.
Meanwhile, my experience with SSDs has been 100% perfection (although I hear they suddenly stop working when they go). And SD / microSD cards are usually good, although I’ve had a few of those die in a few different ways.
Yeah, I also have a (now very old memory-starved) Synology NAS (4-bay). It used to have Green drives, but I’ve switched to (even larger) Red (NAS-specific) ones when each Green died. I’ve disabled everything on it that I could (so the disks could enter sleep mode when I’m not using the NAS) and use an external ARM device for anything not purely file/NAS related.
I actually went through a total of 6 green drives, as 2 died immediately (and I got 1:1 replacements). The first 4 were from 2 different vendors. Not so much later, I had another green die. Everything was good for a while, then I had another go. And then, finally, I had those last 2 die back-to-back, with the second in the RAID rebuild. (As mentioned above.) That’s when I realized that one of the Synology updates re-enabled indexing and also bloated memory usage, thus causing my NAS to hit the drive from time to time and even enter swap state once in a while. The disks couldn’t spin down and sleep long enough and would spin up from time to time. I fixed it right after, and it’s been smooth sailing since (with the 4 WD Reds).
I would also go the DIY NAS route in the future, probably with a btrfs-based RAID. This next time with 1-disk failover and a separate backup snapshot (instead of 2-disk failover). Advantages: Fully Free Software that I can update to the latest kernel, no proprietary lock-in features. Disadvantages: No easy desktop click-through UI for everything with some hand-holding. However, as I happen to work on https://cockpit-project.org/ I’d use that for a nice UI instead, and it does have a disk UI too.
If you’re buying new disks now or in the future, the WD Red drives are now SMR, which are not suitable for RAID. The Red Plus/Pro drives are the regular spinning platter drivers. Seems a lot of people are now buying Segate NAS drives.
