Power on, Power off

[tp4tissue]

Obviously bad for harddrives.

But bad for the PC ?

What say you Xpurts ?

[Leslieann]

Hasn't really been an issue in decades, hard drives included.
So long as your system stays relatively cool being on or off does almost nothing, being on all the time however will take a toll on fans over time.

What kills parts is power fluctuation (especially modern low power parts), under/over power, and for hard drives, vibration.
Really stable power will GREATLY enhance the lifespan of capacitors, crappy power destroys them over time by draining the caps, this leads to power problems further down the line. This is why undervolting can actually be bad if taken to extremes and also why a power regulating UPS is so beneficial to the life of a system and improve overclocks.

As for drives, not only does power matter but old drives haaaaaaate vibration, not bumping, not typing, just loud noise is enough to cause problems with drives, now imagine it falling or being kicked. If you haven't seen it, there's a video on Youtube called "Shouting in the Datacenter" (below), it shows how much your drives can sense even in a noisy room (I love this video). After watching it think about how you treated your laptops or how many times you've kicked or bumped your tower with your chair over the years... I was mostly using SSDs by the time I first saw this but I started being far more protective about my remaining drives after seeing it.

[noisyturtle]

My computer has essentially been on for 10 years straight, barring the 3 times I moved and the 2 times a year I blow it out. Never had a failed hdd. Failed just about every other part, but thankfully never a hdd.

I really ought to make a backup... been putting that off for 10 years too

[tp4tissue]

bits flip in ~5 years, so, it's not just back up, you gotta in and out.

[Leslieann]

If I remember right, by year 5 only about 25% of drives are even still functional, the good news is if it lasts 5 years it's probably going to last several more but you should be regularly checking S.M.A.R.T. stats by year 3. The more full the more often you should check. More free space means more areas to use before it dies. ANY uncorrected errors (read raw data) means it's failing, doesn't matter if it's one or 50, it's failing, it's only a matter of how soon but rest assured it is failing.

Never spinning down or never powering on drives, both are equally deadly. On new old stock the lube goes bad, on drives never spun down the same thing happens. The lube moves to the outside and just cooks over time, tell a network admin to power down servers and you can watch them go pale because they know many drives will simply not come back up. My home file server is the same way, I never spin down the drives because I hate the lag it creates, despite using a home grade drive. This is NOT recommended by drive makers for home grade drives but I haven't really had any issues so long as they stay cool.

In my experience, fans live longer laying flat, the bearings wear uneven when vertical. The exception is ball bearing fans, those prefer to be vertical.
Drives live longer vertically, the bearings are less a problem than the head/arm which is less likely to crash in a vertical orientation. However this is anecdotal, I don't have nearly enough data to say this for certain, however, look at hard drive boxes, server drive orientation and even many full size external drives, they often sit vertical. Considering the risk of falling over you would think drive companies would lay them flat, yet they don't.


BTW
Backup, virus scans and OS install are the hardest your drive will ever work and is when they are most likely to fail, games, since they're downloaded do not load the drive as hard despite writing more data. It's less rare that they fail on OS install but it does happen (even on brand new drives) and can even happen on a refresh. I've seen MANY fail during virus scans or while trying to do a backup on an already failing drive. My advice, before doing a backup or virus scan take a look at the S.M.A.R.T. stats then decide if you want to risk scanning without making a backup first. If it shows bad sectors skip the scan and start backing up important data first, then start a full backup, the reason being you may only get one shot at the data. I usually start the backup process before I even call the customer because I don't want it to have a chance to even spin down, again, it may be the only chance to recover the data.

[tp4tissue]

Let's clarify , when you say drives prefer vertical, do you mean Pancake or Shark fin. ?

[suicidal_orange]

Let's clarify , when you say drives prefer vertical, do you mean Pancake or Shark fin. ?

Considering the risk of falling over you would think drive companies would lay them flat, yet they don't.

Pancakes would struggle to fall over I think

Great info as always Leslieann, I had to laugh at the noise sensitivity video given the guy was shouting just to be heard over the background fans.  Do drives have selective hearing?!

My spinner has survived 4 years of at least once daily spin-up/down, it's 90% full so when capacity runs out it will be replaced.  If that's too long I will spend a long time re-re-ripping my CD collection, just like last time.

[Leslieann]

Let's clarify , when you say drives prefer vertical, do you mean Pancake or Shark fin. ?

Vertical, as is shark fin.

Great info as always Leslieann, I had to laugh at the noise sensitivity video given the guy was shouting just to be heard over the background fans.  Do drives have selective hearing?!

Thanks
Not selective, those poor drives are constantly being bombarded and transfer rates are always lower as a result, as if transfer rates were not already an issue in the server space.

[tp4tissue]

My drives always made more noise vertical. 

Does it have a angular preference ?, ports facing which clock # ?

[Leslieann]

My drives always made more noise vertical. 

Does it have a angular preference ?, ports facing which clock # ?

The arm when it parks will have slightly more force when parking depending on orientation, but it depends on the drive, some park towards the ports, some park to the side and which side depends on the drive. Of course that only effects parking, it could also just be the panel it's mounted to, you aren't mounting it to the same panel the same way. It also may be mounted with different hardware or even just as simple as how the noise is projected towards you.

[fanpeople]

Oh ****

I just checked crystaldisk.

Power on time HDD 1: 2624 days/about 7 years.
                     HDD 2: 1858 days/about 5 years.

YOLO they got another 5 in them.

[phinix]

I remember days when I had to run PARK before shutting down my PC back in 80/90s

So how are those USB drives? I use them for backups and for movies storage - I keep plug them in and pull out without "safety plugging" - am I risking failure?

[suicidal_orange]

So how are those USB drives? I use them for backups and for movies storage - I keep plug them in and pull out without "safety plugging" - am I risking failure?

Flash drives?  Physically they don't care and wont die but the filesystem will corrupt if a file is partially written.  Apparently in Windows 10 they changed the default mode to write immediately rather than cache because people can't be trusted to do it properly...

[phinix]

So how are those USB drives? I use them for backups and for movies storage - I keep plug them in and pull out without "safety plugging" - am I risking failure?

Flash drives?  Physically they don't care and wont die but the filesystem will corrupt if a file is partially written.  Apparently in Windows 10 they changed the default mode to write immediately rather than cache because people can't be trusted to do it properly...

No, spin hard drives, I use those 2.5" USB hdds.

[Leslieann]

So how are those USB drives? I use them for backups and for movies storage - I keep plug them in and pull out without "safety plugging" - am I risking failure?

That won't hurt the drive, usually.
I rarely used it on Windows until I started working with Mac and Linux, now I use almost every time.

Most often what happens is data loss, "safely remove" allows any data being written to finish before you unplug it. Windows is pretty good about being accurate on when it finishes but not all programs or operating systems are and you could yank it before it's done.

However, depending on the drive and cache/memory, removing a drive while being written can go so far as to DESTROY THE DRIVE. This effects flash memory such as SDcards, usually it results in scrambled data, a reformat will fix it but on early SSDs it was known to corrupt the firmware and destroy the drive. It was well known in the industry but never told to customers. Yanking drives during a write is also hard on any capacitors installed on it, this is usually not an instant death but is one of, if not the most common cause of failure on USB sticks/thumbdrives (whatever you call them). Which by the way are EXTREMELY unreliable for backups, if you use usb sticks for backup, have at bare minimum at least 2 copies because they fail entirely too often.

[Leslieann]

Oh ****

I just checked crystaldisk.

Power on time HDD 1: 2624 days/about 7 years.
                     HDD 2: 1858 days/about 5 years.

YOLO they got another 5 in them.

It's not age that actually kills drives it's just when it starts to happen, the numbers you want to look at is uncorrectable sector count and crc fails.

Uncorrectable sectors are bad spots on the platter itself, the drive will try and work around these but the truth is once they start they tend to keep growing.
CRC fails indicate a low power, onboard cache or bad cable. Most often it's the SATA cable. If it's the cable it can over work the drive trying write corrupt data over and over and failing, you may lose data but that's all. It can take months/years to find and eliminate all of the damaged files (dump all you can and start fresh!). If it's not the cable the drive is hosed.

Unfortunately, you could replace them to prevent failure but the new drives could fail even sooner.

[fanpeople]

the numbers you want to look at is uncorrectable sector count and crc fails.

Should I be concerned? I have NFI, I just go off the 'Good' health status.

[tp4tissue]

Concerned, no, but those temps, could use a fan.

[fanpeople]

Concerned, no, but those temps, could use a fan.

It is usually lower but I just turned the AC (only use AC when actually in the room) on and it is the middle of summer here. But now that you mention it, maybe I should turn the computer off when not using it as it gets pretty hot.

I have a Caselabs BH7 and the fan size is super small for that part of the case. I have one exhaust fan but it is like 80mm or similar. I just looked in the window which I have not done for at least a year..... holy **** it looks like the inside of a vacuum cleaner. Should probably clean the computer.

[Leslieann]

You always want to look at raw column because for some stupid reason manufacturers like to use their own method of numbering and not all programs will read it correct (though it is better now than it used to be but SSDs are still the wild west). Being zero'd out is pretty hard to mistake, hence why you want to look at the raw for anything other than zeros as it means something has changed. Keep in mind, names and line numbers are also not universal.


So...
Hard drive 101 (some hard drive expert may disagree with some points of this but it's worked for me for years)
There are spare sectors built onto ever disk, when one is used, it's marked as reallocated (line 05). It's not great to use them but also not terrible because they can be used for multiple reasons, so no need to really freak out if you see some reallocated being used but take it as a warning. Also these may not be permanent, sometimes a format can clear them (sometimes) but to my knowledge not if it's your current boot drive, regardless it's when you run out and/or you start to see uncorrected that you need to worry as your drive is approaching death's doorstep. In this case you have 200 to spare and a SMART warning is supposed to trip when it sees 140 being used, IF SMART monitoring is enabled (it isn't always) and IF it decides to notify you (it can put a popup on your screen but more on this in a moment). In this case reallocated is all zeros hence NONE have been used. Perfect.

A sector can also be marked uncorrectable (line C6) without triggering a reallocated number (I think). In this case none is acceptable according to SMART, however I have seen MANY, MANY times where drives will have multiple uncorrectable sectors and never trigger a SMART warning. I have seen entire sections (many many sectors) of a drive failing as I watch and pull a backup of the data and never trigger a warning so while nice don't rely on SMART to warn you of anything. While I would need several extra hands and feet to count all the drives I've caught mid-failure this way, I can count the warnings I've actually seen on one hand. Once again, this drive shows all zeros in the raw column. Perfect. 

If you see just one of these, seriously, just one, start backing up data.
Do not think it will fix itself (it won't), do not run a scan to see how many more it can find (oh god!), do not even turn off the computer. Disconnect it from the internet to disable updates (or at least turn off updates), try not to even let it sleep, start backing up data NOW. Not tomorrow, not when you have time, NOW. Stay up late if you have to but do it now, at least get the important stuff.  You literally may only get one shot at backing up that data and this may be it. It may die slowly over several months, it could go months without a problem then poof or (and this has been my most common experience) it could go into a cascade failure and wipe out everything as you backup your data. The only thing you know for certain is that it's failing, get your data off of it.


As mentioned the other line to watch is the ultra DMA crc error (line C7), while the SMART data here says the threshold is zero, that's not entirely true. I'm not saying it's good, it means a it misread something or was fed bad data. RF noise, cable crosstalk, quite a few things can trigger this, as mentioned a bad SATA cable is a very common culprit. Point is if you see some numbers here, don't freak, it could simply be a program crashed while writing data, the bummer is you have no idea what data was found to be in error but since it caught it chances are it also re-requested the data. If you see lots or it's increasing try a different Sata cable, if it's low but stable you may never find the cause. This is not a common sign of drive failure but needs to be mentioned because it is a sign of something else usually. Of the few times I have seen the number go crazy, all but two were bad SATA cables, of the two that were not, one had a bad power supply, the other I concluded was the onboard cache had been damaged probably a power surge. Sata cables are cheap and easy to find compared to a new drive. Again, this one shows zero.

I don't think I've ever actually seen a drive with a read error (line 01) or write error (line C8) that didn't also have other more concerning problems.


Far as I can tell this drive is about as healthy as can be, at least as far as SMART data is concerned, just remember it could change at any time and more importantly there is always a chance something not monitored could fail, but as far as smart stats go, this one gets a relatively clean bill of health.