Plausible flash-drive (stick or ssd) maintenance routine

[PDP-8Member]

At the risk of global ridicule, I put a lot of research into this topic which was hard to do, because these days it can be hard to tell who is just a schill, salesman, or someone relying on facts from 30 years ago.

Quickly, it is this: At the ftl (flash translation layer) level beyond the os, the drive’s internal controller (be it usb-stick or ssd) will do garbage collection and wear-leveling on it’s own, *provided* it thinks enough idle-time has passed. And it will, but that’s the key: idle time.

1) Here in this blog, the author does this for usb-sticks by occasionally booting no further than the bios/uefi screen, and letting it sit that way overnight (so the drive doesn’t work with any data) or longer:
https://cgicoffee.com/blog/2021/10/recharge-your-usb-flash-drive-today

Ok, fine. I do that now once in a blue-moon. I don’t have a rigid schedule for it. I’ll put a stick into my 5v usb charger overnight. Seems like no harm is done, but I can’t prove anything, because my stick is not dead to begin with. I’m not trying to revive trash.

2) However, I found this low-key video talking about reviving dead ssd’s with the same technique. Or at least giving it a shot. Only power is connected, not data. Once again, if you listen closely talks about sitting in the bios screen only for anywhere from a half-hour to an hour or more. But what made this seem plausible is the mention that these knackered drive wills get really warm when the controller goes bananas with it’s gc/wear/error-correction routine!

https://www.youtube.com/watch?v=yTZ1dswQTXs

Although he shows old-school separate connections between power and data, I think just sitting in the bios screen will do the same thing.
BUT, perhaps he didn’t leave it in this condition long enough! Just because the drive cooled after an hour may not be the indication that the job is done! I’d give it even more time.

In fact, in a Micron document, they specify 6 to 8 hours for an ssd and they weren’t talking about dead-drive recovery, but merely explaining the process. Is it possible that this tech is not allowing for enough time, albeit getting a quick-fix, and merely restarting a process prematurely that the algo-process actually takes longer, even if the drive has cooled down? Of course no document clearly explains exactly how to ensure enough idle-time to trigger the controller with just power. Sitting in the bios screen for many hours is a neat trick!

3) So I do this now on a blue-moon basis for my sticks. What about drives that aren’t dead? How about just a preventative maintenance test?

I took my windows laptop with an emmc drive, and another windows desktop with an ssd. I erased junk I didn’t need, and manually performed and “optimize” on them. (aka similar to an fstrim), rebooted only into the bios, and let them sit there for 2 hours. After that 2-hour delay, I let it finish the boot.

I noticed a difference! I’ll bet much of that was psychological, but it can’t hurt. 🙂 Makes sense to sit at the bios screen (no data pass), because with Windows, drive activity is nearly always happening! I turn off indexing, but that’s not enough and a different scope.

So, in the end, I may still be a fool, but since merely booting into your bios for say 2 hours once in awhile while the drive is still good, won’t hurt, why not? All this secret stuff was so hard to suss out!

4) I tried this trick sitting in just bios screen with externally connected ssd case via the usb port, BUT the problem is the case will do an automatic shutdown after 30 minutes regardless. However, since I’m not trying to revive trash, 30 minutes of “down time” to allow the ftl controller do it’s thing can’t hurt.

I still feel weird doing this, but at least there *seems* to be some evidence with those unrelated blog post and video kind of backing each other up over this that I don’t feel too foolish.

• This topic was modified 1 year ago by PDP-8. Reason: Added Micron note

[Xunzi_23Member]

Hi pdp-8, good summary.
Powering SSD to put them in to maintenance mode has been posted on several occasions.

The proceedures are not a secret

Actually Pre warranty return requirements.
I picked it up a few
years ago when a crucial 64GB drive went seemingly fully defective.
Power on no data cable connected fixed that drive and later a couple of others for me.

In a big shop in the city there is a stack of drives which customers insisted be
replaced, customer signs a release form for the shop to dispose of the drives
complying with the GDPR,
They get a factory reset with a tool I have not yet been able to find as software,
my assumption is it only gives a couple of commands and the controller clears all
data, it will also update drive software when available.

Software updater may be available for download, depends on drive and all found
up to now are running on windoze..

Early drives with low capacity go in to fail mode much more often than more modern drives
with higher capacity and more reserves.

USB Sticks powering has repaired several which went read only, that is a manufacturer
data protection mode. San disk ultra seems prone to that issue. Two of my four 64GB
drives went that way. Intent is to prevent further corruption a read only lockbit is set,
you can rescue your data before repair, reset may be initiated during the self repair
and maintenance mode.

That gem came from SanDisk.

• This reply was modified 1 year ago by Xunzi_23.

[PDP-8Member]

Thanks Xunxi! You are the first one to tip me off to this, so I’m grateful.

That pushed me into thinking about booting solely into bios/uefi and staying there for a long time was a PM kind of thing. What wait until your disk dies?

Another alternative, at least for usb sticks is to do this – make sure your ports don’t go into powersuspend by adding to the kernel paramaters:

usbcore.autosuspend=-1

Then, you can insert sticks to whatever ports you have available and let them sit there overnight and still be able to use the computer. But I’d be sure to UNmount them if they get mounted automatically. The antiX unmount tool is pretty awesome to make sure.

External ssd’s in their own usb cases might have auto shutdown anyway, so one might have to put them inside the computer proper.

I even do my manual formatting with F2FS, and “over-provision” 20% of the capacity of free-space. But I’m not obsessing though. 🙂

• This reply was modified 1 year ago by PDP-8.

[Xunzi_23Member]

Sadly so much snake oil spread around, buy our super wonderful ssd fixer etc.

Anyways The leave in BIOS method is also a recommendation for M2 SSD.

External SSD, unsure, what makes that more complicated is now both Sata and Msata 1
and M2 can be used in external adapters. Maybe Micro sata too.

Then come the USB SSD Sticks.

Storage Hardware stays interesting 🙂

[PDP-8Member]

Snake oil: You’ll like this. I was cleaning out the back of the drawer, and found to usb 2.0 sticks. And one was made *specifically* by Kingston for the Windows Vista Ready-Boost. Naturally those specific versions cost a lot more.

So I ran a Crystal Diskmark test on it, and was amazed at the difference between it, and another general-purpose one:

Kingston Data Traveler “Ready Boost” How did they program the controller to be so exactly alike in read/write here? We’ll never know.

Read / Write
————
18.2 / 18.2 < Consecutive
18.2 / 18.4 < Slighly harder consecutive
6.7 / 0.1 < Easy random
4.4 / 0.1 < Harder random

Kingston DTSE9

Read / Write
————
29.7 / 6.0
30.38 / 8.8
4.8 / 0.0 < so slow it didn’t even register
4.6 / 0.0

I vaguely remember that the purpose-designed one was actually working to a degree, whereas the random general purpose didn’t see to do a thing, even though one had to purposely create an RB stick out of it first. No wonder. 🙂

• This reply was modified 1 year ago by PDP-8.

[vitforlinuxMember]

I have always thought that everything, but really everything, breaks.

A technician talking about mechanical hard drives said that their lifespan is 6 years, and if there is a teenager at home, 3.

I used two 32-gigabyte USB drives for a few years, and one I used to try out different distros, and it broke first, then a few months later the other one.

Well, they still read. To replace them, I got three 64-gigabyte USB drives, but I will divide the work better.

Important small files on one USB drive that works less.

Older and smaller USB drives seem to handle stress better.

Sorry for my spaghetti english, i'm italian. Happy XORG user. GNOME? No thanks!

[RobinMember]

vitforlinux wrote:

A technician talking about mechanical hard drives said that their lifespan is 6 years, and if there is a teenager at home, 3.

Since you are talking about spinning devices: I have not seen them fail before 10 to 15 years constant everyday usage, some even 24/7/365, teenage at home or not. But maybe that reliability of spinning drives has changed the last years, when the capitalist idea of forcing consumers to buy new stuff constantly by means of well hidden planned obsolescence has become more and more ubiquitous and wearing-out strategies are built in by default in the devices instead of designing them long-life. All under the hood and label of progress, miniaturisation, increased speed and increased packing density. Sad truth is: It is no technical category. This is simply needed in capitalism to keep the economy boiling in hard times.

Windows is like a submarine. Open a window and serious problems will start.

[vitforlinuxMember]

When I said that thing about 6 years, I wasn’t completely in agreement. I used a mechanical hard drive for 4 years after it was replaced, and another one that was half broken… only in the part where Windows was installed.

But a nice backup of important data is always a good thing.

Sorry for my spaghetti english, i'm italian. Happy XORG user. GNOME? No thanks!

[Brian MasinickModerator]

vitforlinux wrote:

When I said that thing about 6 years, I wasn’t completely in agreement. I used a mechanical hard drive for 4 years after it was replaced, and another one that was half broken… only in the part where Windows was installed.

But a nice backup of important data is always a good thing.

I’ve gotten away with using spinning hard drives that were already producing errors; I simply did file system checks and disabled bad blocks and kept going. But I’ve had MORE than a decade of success with spinning devices, several produced well over a decade of solid performance before errors or decline occurred. Also Linux file systems have brought life, ESPECIALLY with antiX, to multiple old systems. I had all of the Dell D600 series units either at home, work, or fix ups for others; adding MX Linux, PCLinuxOS or antiX to them gave them SEVERAL additional years.

I have an HP desktop d5000 series, which was actually a Compaq computer architecture acquired by HP. It’s about done now, but I have a Lenovo Thinkpad X201 only slightly younger that’s still going strong.

--
Brian Masinick
Alternate "Search B":
This search page

[Author]

Posts