Apple SSD Health status >4.2TB, how I can tell it is dead?

Question

I've installed SSD Health (512GB) on my MacBook Pro (late 2013) which shows the following stats:

Normally Health status should show the percentage, but it's showing >4252GB data written.

I've experienced recently some data loss (like copied files are filled with NULLs, SHA256 mismatches while installing packages, etc.), and First Aid found HD corrupt (exit code 8). Does it suggest my SSD is dead and it needs to replaced (if so, based on what information exactly)? On the other hand, shouldn't SSD drives work for at least 10 years?

---

Here are few reproducible data loss examples:

$ wget -q https://homebrew.bintray.com/bottles/glibmm-2.54.1.sierra.bottle.tar.gz
$ ls -la glibmm-2.54.1.sierra.bottle.tar.gz 
-rw-r--r-- 1 kenorb staff 10033965 Sep 18 22:54 glibmm-2.54.1.sierra.bottle.tar.gz
$ gzip -t glibmm-2.54.1.sierra.bottle.tar.gz 
gzip: glibmm-2.54.1.sierra.bottle.tar.gz: unexpected end of file
gzip: glibmm-2.54.1.sierra.bottle.tar.gz: uncompress failed
$ cp -v /Volumes/SSD-256G/file.txt . # Copy file from the external SSD.
'/Volumes/SSD-256G/file.txt' -> './file.txt'
$ diff /Volumes/SSD-256G/file.txt file.txt
Binary files /Volumes/SSD-256G/file.txt and file.txt differ
$ hexdump file.txt
0000000 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
1b86060 00 00 00 00 00 00                              
1b86066

Answer 1

"Copied files are filled with NULLs, SHA256 mismatches" are indeed indicators of possible SSD failure. I would not expect an SSD to live 10 years unless you keep it on the shelf.

If the same file reads differently (e.g. md5 path/to/file yields different hashes) in any two attempts it is sufficient indication that the SSD is terminally ill (of course provided that the file is not modified). You can test this on any large file to start with, the bigger the better.

If that test passes, you can proceed doing the same for the whole disk. You'll need to boot from another bootable medium from which you could calculate the hash of the whole SSD. For example, boot into Linux live USB/CD and run md5sum /dev/sd<your ssd letter> twice — the results must match for a healthy disk.

Finally, if the above tests pass, while you're still booted into Linux, run short and then long test provided by smartctl:

smartctl -t short /dev/sd<your ssd letter> # results can be seen after a few minutes
smartctl -t long /dev/sd<your ssd letter> # a few hours to wait

Answer 2

The Apple's fsck_hfs tool has the -S option to scan disk for bad blocks.

For example:

diskutil list # Note the disk path.
fsck_hfs -S /dev/disk22

This will scan entire disk for https://en.wikipedia.org/wiki/Bad_sector, and tell whether it is corrupted, or the software repair should be enough.

---

To avoid having disk being mounted, the best is to boot the system from either a single mode or macOS installer (USB or DVD) and run Terminal app.

Here are sample commands to repair the encrypted volume (as a root user):

diskutil coreStorage list # Note the UUID.
diskutil coreStorage unlockVolume 11111111-2222-3333-4444-555555555555
diskutil umount /dev/disk22 # Unmount after it was mounted with write access.
fsck_hfs -y /dev/disk22 # Check and repair any errors.
fsck_hfs -S /dev/disk22 # Scan entire disk for the bad blocks.

If errors were repaired successfully, and there are no bad blocks, it can give some confidence the SSD Health is still good. However, if there are still some disk errors and no bad blocks, 3rd party repair software can be used (such as DiskWarrior, it can repair corrupted directories/files and diagnose the hard drive using S.M.A.R.T. diagnostics).