Question about Hard Drive Surface Test

simtra said:

Once a drive starts to report surface errors, it normally is not long until the drive just dies totally. If you think about why there are surface errors (like dirt or scratches) it is normally safe to assume that the problem will show up on other areas of the drive...

Click to expand...

I am not sure I can totally buy that assumption. HDDs all have surface errors from the beginning. The original full scan done before you purchase it sectors out the errors that are there. The surface scan in the DVR does the same thing.

"Repaired" is a misnomer; no potholes are filled by this process. Instead, a map of where the potholes are is kept by the drive and it is instructed to simply not write there. Depending on how the surface scan routine is written, it may be also recounting errors that were sectored out during production along with new errors. This means counting them is a fairly inaccurate path to drawing a conclusion about that.

It is not unusual for surface errors to appear on a HDD after it is in use, but if you are seeing a lot of them (which is hard to judge from the test) it could mean it is failing, yes. There is just no true connection between reported surface errors and a failing drive that is a reliable indication of that.

A new drive has surface errors pretty early on, and will have more as it is used. This explains why on a PC drive files will become fragmented even if there is contiguous space for a file (it writes, sees errors, and continues writing somewhere else). Since drives are built in clean rooms, there is not much chance of dirt of debris, and the head does not ever (hopefully) touch the surface; if it ever did you would have much more severe problems to worry about. So "scratches" is also pretty much a misnomer.

Surface errors, once the original ones are sectored out, are the surface flaking off. Yes, that causes debris, but the flakes are small enough to not get between the surface and the head, and are usually forced to the edge of the housing just due to the centrifugal force of the air movement over the surface, meaning the odds of minor flaking causing more problems are small, and major flaking means you already have a serious issue anyway. The flaking is aggravated by expansion and contraction, which is one of the reasons why heat is not good for longevity, and also contributes to the theory that a drive that spins up and down will fail earlier than one that spins continuously (which is of course just the opposite if it is the bearing that fails).

The correlation is not between surface errors and failure, as much as it is between suspecting HDD problems and finding them (suspected drives are often bad drives, while those not suspected often are not). The surface errors it finds may or may not have anything to do with the actual health of the drive; you would have to know how many any what that amount actually means to draw a correlation.

P Smith said:

Sometimes bad power (noise, ripples, glitches) could produce 'soft' error during writing. There is methods to erase internal "G" list of bad sectors and run new scan with remapping what would recover these 'soft' bad sectors, also it would find and remap really weak/bad sectors to spare area. So far I saw in Seagates the limit is 4096.
About DVR utils - they are pretty easy and while developers keep silence about its algos, judging by results and spent time I would say these are not that comprehensive as should be.

Click to expand...

IME, "bad power" will most definitely reduce the lifespan of hard drives. All my computers are hooked up to APC UPS, and 2 of 3 HR24 DVRs are on UPS also. We live in an area of Los Angeles with usually good and reliable power, but occasionally during a summer heatwave or winter T-storms, the UPS will start beeping to indicate a power condition (brown out I suspect) and about 50% of the time, we loose power after the UPS has "warned us".

P Smith said:

If you could take out the drive and run some program (MHDD from a CD), then we could know its health [SMART] and by scanning - how many bad sectors it will develop by one pass.

Click to expand...

One of the biggest advantages to running an external eSata setup is the ability to take it off the HR and put it on a PC for diagnostics. Lately I use the Western Digital software. At an old job, we used to use software called SpinRite, and on the deepest level of testing it claimed it could test and return good sectors back into use that had been marked bad. I don't recall seeing this happen, but I do recall many times it would test and remove bad sectors, and if it was more than a handful, we would replace that hard drive. If you smell smoke, there is probably a fire nearby.

nsolot said:

IME, "bad power" will most definitely reduce the lifespan of hard drives. All my computers are hooked up to APC UPS...

Click to expand...

Can't imagine why. It's not as if the drive starts to dig the actuator into the platters when the voltage drops or anything else that would portend a catastrophic failure. About all that happens when a HDD loses power unexpectedly is that part of whatever data is in the buffer can spew across adjacent sectors as the actuator arm moves where it shouldn't. That will not cause disk errors, but it could corrupt data, and if that data is part of the OS, buh-bye OS.

The voltage regulator in modern power supplies usually keeps the DC at proper levels even if AC fluctuates somewhat. If AC falls enough to cause the HDD to not be able to spin at proper speed, another problem happens that is worse, which is that DC ripple increases, and that can corrupt data as read or written for that point in time. A UPS can come in handy for that, but again, this will not damage the drive, only the existing data written on the drive.

DC ripple can also happen as a common soft failure of a PS, and this can cause random reboots for the same above reason. No UPS in the world can help you then, because DC ripple is generated downstream of the AC power.

...we used to use software called SpinRite, and on the deepest level of testing it claimed it could test and return good sectors back into use that had been marked bad. I don't recall seeing this happen, but I do recall many times it would test and remove bad sectors...

Click to expand...

SpinRite does not work that way.

When your PC tries to read the HDD, it tries over and over again until it reads the data (some data is harder to read if there are soft errors in the magnetic medium). It usually tries about 10 times or so, and if it hasn't got it by then, it gives up.

SpinRite does the exact same thing, except it virtually "never" gives up. It will keep trying for days to read data from bad sectors before giving up, and if if is successful it moves that data to a known good area on the drive. Then it moves on to the next sector and the next until it finds another problem, and keeps repeating that process until it has read every sector it can read on the drive that has data on it.

But again, there is no actual repair of the medium going on here. There is a DB with a map of where the known bad sectors are, and the drive uses that map to avoid those sectors. It's like a permanent no-fly zone for the actuator. SpinRite recreates and updates that DB. It's just that simple. No repair, no magic.