Which S.M.A.R.T. stats are the best indicator of an imminent HDD failure

[nic2k4]

Backblaze (one of the largest data center operator) has an interesting post on the subject.

[redwudz]

That all does sound reasonable. But for the drives I've dealt with over the years; When I get a S.M.A.R.T. notification, I get my important data off that drive.

I sometimes have trusted the drive for minor uses afterwards, but nothing critical. In almost every circumstance, the S.M.A.R.T. errored drive really did fail not far in the future.

IMO, HDDs are cheap. Toss the drive if you get a S.M.A.R.T. error and don't take any chances.

[Bjs]

Raw read and spin up time ... when these values drop to 80% it's time to move data elsewhere.

[fritzi93]

Raw read and spin up time ... when these values drop to 80% it's time to move data elsewhere.

That makes more sense to me as far as the home user is concerned. Data center usage is a very different scenario, no?

[nic2k4]

Data center usage is a very different scenario, no?

I would think so as well, at least when start/stop count and power on hours goes. Like they say in the post their drives have few start/stop and long power on hours, which is likely the opposite of most home user. It's probable that a combination of those two value would be indicative of a pending failure. One thing is for sure, the closer the power on hours gets to the drives published MTBF the more you should expect problems.

I monitor my drives with Crystaldiskinfo, there are four settings that they seem to think matter (they're the only ones you can reset the alarm threshold):

Reallocated sector count
Current sector pending count
Uncorrectable sector count
Remaining life

I wonder if the Backblaze report will change anything.

[hello_hello]

I have a PC with a pair of old 320GB WD drives running as a RAID-0 volume. Power on hours is around 4.2 years for both drives. Due to reading this thread I looked at the drives with SpeedFan a few days ago and discovered the sector pending count wasn't zero for one of the drives. Crystaldiskinfo looked at the S.M.A.R.T. info and rated the drive as "caution". Two days ago the PC blue screened and when I rebooted, the drive in question didn't appear in the RAID BIOS. When it did, the PC still wouldn't boot from that RAID volume. I left the PC off for a few hours and tried again. It booted fine and I successfully copied the files I hadn't backed up to an external drive, then made an image of the C partition so if the worst happens I can swap out the drive and be back where I started pretty quickly. Sector pending count has increased a little. The PC's been running again for over 24 hours without an issue. It'll be interesting to see how much longer it keeps running.

I have no doubt S.M.A.R.T. warnings are probably something to be taken seriously, when they happen, but I've had more drives die without S.M.A.R.T. indicating there's anything wrong, even while the drive is making alarming noises and moving data at a snail's pace, so I don't tend to worry about monitoring S.M.A.R.T. info.

I can't complain too much. Aside from a couple of drives I dropped and buying a pair of Seagate drives with known issues (returned under warranty several times) I've had a pretty good run with drives over the last five or six years. The PC I referred to above also has a pair of Samsung 500GB drives with a total 4.2 year power on time and they're still going strong. The 2TB Hitachi drive connected to this PC via a USB dock has a power on time of about 3.8 years and it's still working fine. The four drives I removed from this PC a while ago (replaced with larger drives) had a total power on time of 4.5 years each, if memory serves me correctly, and they're working in semi-retirement as external backup drives.

[nic2k4]

I would suggest you remove the control board from the drive that's having issues and wipe the pads with a white pencil eraser (Staedtler type), that should extend its life. I've brought drives back to life that way.

[newpball]

SMART or not after about a year of heavy use every drive I have goes in the "unreliable, good for extra backup bin".