Skip slow blocks quickly

How, what, where and why - when using the software.

Skip slow blocks quickly

Postby QuantumX » 2021.07.22. 08:58

Good day,

I'm running a surface test on a disk that is having problems.

The idea is to determine which sections of the disk is unaffected and split the disk into sectioned partitions that use the unaffected areas.

The problem is that during surface test the test gets stuck for more than one hour on a single weak block.

Considering with this specific disk one block indicates about 190MB, that means in one hour the test was able to read at only 0.052MB/s.

At this rate if only 10GB of this 2TB drive is problematic the surface test would run in the area of 60 hours.

Is there a way to set a threshold that marks a block as damaged if the speed the block is read is below a certain amount?

For example, the disk in question normally has a read speed of 110MB/s. In my mind any block that reads slower than 55MB/s is already "compromised", how much slower than 55MB/s I don't care since it already failed in my mind and I would prefer to exclude it when partitioning.

Would be nice then if I could set 55MB/s as the cutoff speed and then if a 190MB block takes longer than 3.5 seconds to read (190MB / 55) it will be marked as damaged and the test moves on to the next block?
QuantumX
 
Posts: 4
Joined: 2021.07.22. 08:40

Re: Skip slow blocks quickly

Postby hdsentinel » 2021.07.22. 10:02

Generally when problematic sector(s) found, Hard Disk Sentinel automatically performs retries and in this case, probably the retries cause the very slow speed (especially if there are high number of problems in the appropriate block).

You can reduce the number of retries during the test to skip problems sooner: please select Disk menu -> Surface test and select the drive+test type but before starting the test, select the Configuration tab. In the bottom right area, you can reduce the maximum try count, so you can select 1 (or even 0) to skip problems faster.

All blocks with damaged (unreadable) sectors will be displayed as RED.
All blocks with no damage but slower speed are displayed with darker green blocks to indicate the slowness: darkest color means slowest areas.

By this, you can determine which area(s) of the disk drive should be excluded from a partition.
For example this page: https://www.hdsentinel.com/kb/category/4/disk-testing-and-repairing/darker-green-blocks-during-disk-testing.html
shows a typical situation when the end of the disk drive contains much slower (darker green) and unreadable (red) blocks.

By moving the mouse over the appopriate block on the disk surface, you can read the sector number and MB position on the bottom as this information helps to create partition(s) outside such problematic range.
User avatar
hdsentinel
Site Admin
 
Posts: 2597
Joined: 2008.07.27. 17:00
Location: Hungary

Re: Skip slow blocks quickly

Postby QuantumX » 2021.07.22. 10:20

Sorry, I forgot to mention this slow scan speed is with retry count already set at 0

The problem is the disk only reads a small bit of data every few seconds which makes it extremely slow even before retrying.

Image

Currently the scan did 264 blocks (around 50GB) in 2h 15min so this is looking like it might be a 90 hour scan.

If setting the retry count at 0 is the only possibility then I suppose you do not have a solution for me?
QuantumX
 
Posts: 4
Joined: 2021.07.22. 08:40

Re: Skip slow blocks quickly

Postby hdsentinel » 2021.07.22. 11:55

If the beginning of the drive is so slow (if only 264 blocks processed, so 2.64% completed) then the beginning of the drive should be completely ignored in a partition.

So I'd try the following (after stopping the test, as no real reason to continue this way):
select Disk menu -> Surface test (and the appropriate options, eg. try count = 0) and select Sequential backward test on the Configuration page and unselect the default Sequential test. Then start the disk test.

This way the test will start from the "end" of the surface and advance to the beginning. Ideally it will perform at normal speed and somewhere (ideally near block 300-400) it will reach the problematic slow/damaged area. Then you can create a partition to cover this area and a big partition to the remaining capacity and then you can delete the previous small partition as it is no longer required and use only the remaining capacity of the disk drive.
User avatar
hdsentinel
Site Admin
 
Posts: 2597
Joined: 2008.07.27. 17:00
Location: Hungary

Re: Skip slow blocks quickly

Postby QuantumX » 2021.07.22. 12:34

Sorry again for not providing all the details in my initial post. The scan is running with the randomized option.

So far the problematic blocks seems to be towards the end of the disk but there are also dark green blocks in the middle.

Would this make sense as a feature request to allow the user to set damaged block = very slow block and then be able to input the threshold for what is "very slow"? Or have a 3rd block type called "SLOW" mapped in dark gray or something with an option to specify what qualifies as "SLOW" in KB/s or MB/s and such block then be skipped quickly during read test? Maybe call this "accelerated scan" or something

Code: Select all
if("block size in MB" / "elapsed read time" <= "threshold set by user"){
    "Mark block as SLOW";
    "Skip to next block";
    }
QuantumX
 
Posts: 4
Joined: 2021.07.22. 08:40

Re: Skip slow blocks quickly

Postby hdsentinel » 2021.07.22. 14:37

Sorry, but generally things are working exactly this way (except the "dark gray" color): when the block is slower than should - then it is displayed with very dark green color. A very slow block is displayed with very dark green (almost black).

Of course we can determine the block speed after the block is tested (completed) so only then can determine the actual color and only then can advance to the next block.
This is exactly how things work now...
User avatar
hdsentinel
Site Admin
 
Posts: 2597
Joined: 2008.07.27. 17:00
Location: Hungary

Re: Skip slow blocks quickly

Postby QuantumX » 2021.07.22. 15:04

Ok, thanks for confirming. I will let the scan run for a few days then
QuantumX
 
Posts: 4
Joined: 2021.07.22. 08:40


Return to Questions

Who is online

Users browsing this forum: No registered users