Really weird SSD issue

SiriusStarr · #1 29th June 2011, 07:30 AM

Hi all,

I'm having one of the most perplexing issues I've ever encountered and can't seem to find any record of anyone having similar issues (though it's possible my search engine skills are failing me), so was wondering if anyone had any input.

I run Fedora 15 on a 128 GB Kingston SSD and have had no issues up to this point. Today, I installed some software (just libreoffice-base, nothing system related) and it seemed to install fine, but upon attempting to open libreoffice was getting a weird error message. Upon trying to run it a few more times, Java was crashing as detected by ABRT, so to eliminate the possibility that something was just borked, I went for the Windows approach and rebooted.

Upon rebooting, however, I was treated (consistently) to a ramdisk: inconsistent write kernel panic. Figuring something was messed up in the kernel/bootloader, I attempted to get into my bootloader options, but it would not accept my bootloader password, strangely. I then booted with the install CD into recovery mode and went into grub.conf and deleted my bootloader password. Upon rebooting, however, the bootloader was still asking for a password. Booting into recovery mode once more revealed that grub.conf appeared to be unaltered by my previous edit. I then pulled the drive out and stuck it in an external enclosure and accessed it with another computer, bringing us to the bottom line:

I apparently cannot permanently write to my SSD. With the drive mounted, I can view and edit files normally, and the changes appear to be persistent. However, upon unmounting/remounting, all writes have been reversed. This is probably the root cause of all of the above mentioned software and kernel panic issues.

My question is this: How do I go about figuring out whether this is a hardware problem or something else? I don't think it's a filesystem issue; the problem occurs on all three partitions that were on the drive and fsck reveals no problems. Has anyone heard of an issue like this, and is it likely that my SSD is just dead and needs to be RMA'd?

Many thanks,
Sirius

---------- Post added at 11:30 PM ---------- Previous post was at 10:31 PM ----------

Okay, the issue is definitely hardware related... The drive will not even format. Any attempt to format "completes successfully" but does not actually affect the disk... Any thoughts are still appreciated, however.

rtguille · #2 30th June 2011, 01:07 AM

> I've ever encountered and can't seem to find any record of anyone having similar issues

Definitively, it is a weird issue. I also never imagined this can happen... it is like having an "ultimate snapshot" drive.

have you tried contacting vendor support?

---------- Post added at 09:07 PM ---------- Previous post was at 09:02 PM ----------

Check kingston support for firmware updates and/or call support.

http://www.kingston.com/latam/suppor...ow/default.asp

There are some "urgent" firmware updates for some models, check yours.
(also, backup your data)

SiriusStarr · #3 30th June 2011, 03:59 AM

I'm fairly certain the drive itself is hosed. I'm going to contact Kingston for a replacement. Tried overwriting with zeros with "dd" and that didn't work, so just going to try an ATA secure erase. And if that fails, then it's RMA. Poking around on the net, it seems like these drives just have issues, so hopefully they'll send me something new. It's still under warranty, thankfully.

stevea · #4 3rd July 2011, 07:26 AM

Were the file systems on the SSD mounted with the 'discard' option ? Otherwise the garbage collection can get terrible. Shouldn't fail but .....

IMO you should select any SSD drive based on the controller chipset used. At this current time I'd strongly opt for a Sandforce 1200 or better, or an Intel or Crucial (Micron Tech) proprietary.

You might try the 'hdparm --trim-sector-ranges ...' option *AFTER to take any backups.

SiriusStarr · #5 3rd July 2011, 07:31 AM

Yes, TRIM was enabled on the drive. The controller itself seems to have died. I did manage to erase it with an ATA secure erase, but it still doesn't work, so I'll RMA it.

I would have picked a better drive had I been terribly concerned, but this one was quite cheap and was better than the HDD that I'd previously been using (and am now perhaps ironically using again). Thanks for the input on controllers though. :-)