Lost all data on home partition, would really appreciate advice

litemotiv · 2008-08-11 21:15:35

since recently i have an ssd in my laptop, i have my /home on sda4 and use noatime and nodiratime for mounting (was advised to do so elsewhere).

just now i rebooted and got a report that some directory errors were found, a check was forced. about 10 errors were found and corrected, all about a dircount being 1 too high (something like found=5, should be 4, fixing). after this the computer booted through fine, though i soon found out that my /home partition is now completely empty.

what can i do? did fsck wrongfully fix things that weren't necessarily wrong? since the fixes it did seem so 'trivial' you'd think that most of the actual data should still be there...?

any help is greatly appreciated,
-lite

Last edited by litemotiv (2008-08-11 21:16:37)

litemotiv · 2008-08-11 21:48:40

p.s., df shows that 38% of sda4 is in use, so i figure the data is still really there. i'm no expert, but i would suspect that the directory index was 'fixed' based on erroneous data in the journal (noatime/nodiratime's fault?)..

litemotiv · 2008-08-12 12:20:19

To add some use to this thread, SSD's on linux obviously seem to be a risky business. I've posted a description of my problems on the OCZ support forum (i have an OCZ Core64 SSD), hoping they can provide any more advice or feedback on how to use them properly with Linux. Since more and more people are starting to use SSD's (and might experience problems with them sooner or later), i'll post any info that comes up here for further insight.

It might be a good idea to rename this thread (moderator?) as well, since i'd rather invest my time in finding the cause(s) of the problem than in recovering thousands of scattered lost files.

Here's a link to the OCZ support topic: http://www.ocztechnologyforum.com/forum … p?p=289049

And here's a copy of the opening post:
---

Since 2 weeks i have an OCZ Core64 in my Dell XPS M1330 laptop, running Linux kernel 2.6.25. In those 2 weeks the filesystem got seriously corrupted twice, and i really need to know what's going on before taking any more risks with my personal data.

[General]
Latest (A12) bios, AHCI off

[1st run]
Filesystem: JFS
Writecaching: Off
Options: elevator=noop, noatime, nodiratime
Suspend/hibernate: Off
Performance: Very good

Result: after 2 days, an error message was displayed during boot, saying "init not found". i then found out my complete root-partition was gone.

[2nd run]
Filesystem: Ext3
Writecaching: Off
Options: noatime
Suspend/hibernate: Off
Performance: Good/very good

Result: after a week during which several filesystem checks failed with minor (correctable) errors, a message was displayed during boot saying that directory errors were found that needed to be fixed. Consequently, around 15 errors saying "directory count = 9 should be 8, fixing" were given after which the system booted through fine. I then discovered my entire /home partition was empty, it's contents (20GB) moved to countless anonymous entries in Lost+Found.

I would really appreciate any insight in the potential cause(s) of these problems, and advice on how to avoid any further data corruption of this magnitude.

Note: Dell also sells this laptop with an SSD preinstalled and Ubuntu Linux as operating system.

arch0r · 2008-08-12 12:29:22

avoid using noatime! every time i used it, my system freezed. afterwards i killed it and the filesystems were badly corrupted

litemotiv · 2008-08-12 12:54:04

arch0r wrote:

avoid using noatime! every time i used it, my system freezed. afterwards i killed it and the filesystems were badly corrupted

seriously? darn, that's about the first thing every ssd faq/wiki/etc says you should use...

Procyon · 2008-08-12 13:36:40

arch0r wrote:

avoid using noatime! every time i used it, my system freezed. afterwards i killed it and the filesystems were badly corrupted

Does that go for non-ssd as well?

Ramses de Norre · 2008-08-12 13:47:23

I use noatime on several partitions (all on regular hard disks) and never had problems doing so.

litemotiv · 2008-08-12 14:05:06

i have a feeling my problems are caused by incomplete writebacks on the moment the computer reboots/halts. strange though, since as far as i can tell all disk caching is disabled. the major crashes however all seem to have been after a quick save-all & reboot sequence (i think).

freakcode · 2008-08-13 01:43:52

Nothing in /home/lost+found ?

litemotiv · 2008-08-13 10:44:15

freakcode wrote:

Nothing in /home/lost+found ?

yes, everything (20GB) was in lost+found, but as thousands of anonymous inode numbers.

i just had another error-run as well:

[3rd run - all default]
Filesystem: Ext3
Writecaching: Off
Options: -
Suspend/hibernate: Off
Performance: Good

Result: after one day a bootmessage is displayed saying that errors are found on the root-partition which need to be fixed. Fsck finds and corrects literally hundreds of errors, after which the system restarts and boots through. Lost+found contains around 90 items this time, the boot process claims vboxdrv cannot be found.

edit: vboxdrv not found error might be due to kernel upgrade, non related

Most of the errors found were about Dtime and imagic flags on inodes, although there were some more cryptic ones like Special (device/socket/fifo) inode 486932 has non-zero size.

Last edited by litemotiv (2008-08-13 10:48:33)

freakcode · 2008-08-13 18:22:27

litemotiv wrote:

Most of the errors found were about Dtime and imagic flags on inodes, although there were some more cryptic ones like Special (device/socket/fifo) inode 486932 has non-zero size.

I had dmesg outputing errors like that (XXX has non-zero side) once, with an old (and probably bad) IDE HD.
The disk being SSD shouldn't be a problem, even with atime enabled, as the lifespan is reasonable enough, and for SSD there's no risk of corruption if the access is serial or random. Also, you tried 2 different filesystems (JFS and EXT3), and both support journaling.

If those problems are not being caused by incorrect mount/umount options, I would say it looks like bad hardware.

litemotiv · 2008-08-13 21:49:30

freakcode wrote:

If those problems are not being caused by incorrect mount/umount options, I would say it looks like bad hardware.

Yes so would i.

I just received a first reply from an OCZ support moderator however, and he's sort of looking the other way (not surprisingly maybe).

OCZ support wrote:

I have been running the 64GB here on my Dell laptop (965 / ICH8-M) for 2 days now.. still trying to see if I can generate errors, but I am running Vista.
I have a feeling that unless there is a specific driver for the OS, we may not be able to solve the issues.

And my reply:

litemotiv wrote:

As far as i know Intel doesn't provide specific drivers for linux, so that would mean the Core needs a big "Warning: don't use on linux" sticker? No offense obviously, but since there are so many people using or switching to linux these days (Ubuntu is really becoming a mainstream OS fast) a structural compatibility problem with linux would be a serious issue.
I assume that the important questions to be answered at this stage would be:
- how can we rule out this is not a problem specific to my drive?
- are there any specific directions from the OCZ technical staff for the use of this drive on non-Windows computers?
- is the official OCZ standpoint that custom drivers are needed for any OS running this drive, in which case the warning label above should seriously be issued?

I'll post any updates here.

freakcode · 2008-08-14 13:38:48

litemotiv wrote:

freakcode wrote:
If those problems are not being caused by incorrect mount/umount options, I would say it looks like bad hardware.
Yes so would i.
I just received a first reply from an OCZ support moderator however, and he's sort of looking the other way (not surprisingly maybe).
OCZ support wrote:
I have been running the 64GB here on my Dell laptop (965 / ICH8-M) for 2 days now.. still trying to see if I can generate errors, but I am running Vista.
I have a feeling that unless there is a specific driver for the OS, we may not be able to solve the issues.
And my reply:
litemotiv wrote:
As far as i know Intel doesn't provide specific drivers for linux, so that would mean the Core needs a big "Warning: don't use on linux" sticker? No offense obviously, but since there are so many people using or switching to linux these days (Ubuntu is really becoming a mainstream OS fast) a structural compatibility problem with linux would be a serious issue.
I assume that the important questions to be answered at this stage would be:
- how can we rule out this is not a problem specific to my drive?
- are there any specific directions from the OCZ technical staff for the use of this drive on non-Windows computers?
- is the official OCZ standpoint that custom drivers are needed for any OS running this drive, in which case the warning label above should seriously be issued?
I'll post any updates here.

That's BS. Doesn't matter if the device uses a magnetic plate or solid state internaly, as long as the interface and controller are the same (SATA), drivers shouldn't really matter - unless they implementation is so f**** up that it needs some sort of driver hack to avoid data corruption.

But what I meant with bad hardware was that you could have got a defectuous drive.

litemotiv · 2008-08-14 18:27:01

freakcode wrote:

That's BS. Doesn't matter if the device uses a magnetic plate or solid state internaly, as long as the interface and controller are the same (SATA), drivers shouldn't really matter - unless they implementation is so f**** up that it needs some sort of driver hack to avoid data corruption.
But what I meant with bad hardware was that you could have got a defectuous drive.

Yes absolutely, i agree.

OCZ service rep. no#2 stepped into the discussion today:

OCZ support person 2 wrote:

Im pretty confident this is a driver issue or a bios controller issue. SSD is not and should not be treated like a normal HDD, the way files are written to it and read from it is not the same as a standard HDD, so I am thinking drivers made for standard HDD may be the issue here.

*sigh*

KimTjik · 2008-08-14 19:25:12

That was I really embarrassing statement by the OCZ representative. Too bad since OCZ folks traditionally have been good in giving support for their RAM and OC-settings. This though looks like some need to go back to school.

wonder · 2008-08-14 19:39:50

why do you use ext3 over ext2? ext3 is a journaled file system and i believe that is not suggested to use that type of file system because you ruin fast your SSD drive

litemotiv · 2008-08-14 20:00:11

wonder wrote:

why do you use ext3 over ext2? ext3 is a journaled file system and i believe that is not suggested to use that type of file system because you ruin fast your SSD drive

I wanted to use as much a "stock" linux install as possible, meaning a scenario that millions of others users could also face. Ext3 might shorten the lifespan of the drive in the longer run, but for testing ext3 "should" be one of the filesystems with the least amount of actual data corruption/loss.

jbooth · 2008-08-14 21:15:49

I haven't had any problems with ext2 on my eee pc's 4 gig or 16 gig internal drives.

I have previously had problems with DoA OCZ ram and no longer buy from them after a resolved-to-my-satisfaction but overall poor support case.

Of note: The OCZ guys may (sorta) know what they're talking about for "drivers" -- they probably have a special disk driver that knows about flash memory and how not to wear it out. It is likely their equivilant of a noatime, non-journaled filesystem. Problem is they've dumbed it down beyond the level of usefulness in an attempt to reach "Joe Q. Public"

Last edited by jbooth (2008-08-14 21:18:21)

freakcode · 2008-08-15 01:01:16

wonder wrote:

why do you use ext3 over ext2? ext3 is a journaled file system and i believe that is not suggested to use that type of file system because you ruin fast your SSD drive

I repeat: if maintaining a journal is enough to ruin the data within 2 days, something smells bad on that SSD implementation. The filesystem shouldn't be the cause. I mean, why would I want a next-gen SSD drive that requires software magic (drivers), bypassing a common SATA drive behaviour, to avoid data corruption? Doesn't sound reliable to me. But anyway, I'm now interested and doing more research about SSD drives.

Last edited by freakcode (2008-08-15 01:02:15)

rooloo · 2008-08-15 10:52:15

Im pretty confident this is a driver issue or a bios controller issue. SSD is not and should not be treated like a normal HDD, the way files are written to it and read from it is not the same as a standard HDD, so I am thinking drivers made for standard HDD may be the issue here.

And you shall reply with " I'm pretty confident as well unless OCZ steps up and fixes my issue with this drive this will be the last time I ever buy or recommend anyone to OCZ." I don't have any respect for a company that passes blame.

BTW, just to be a smart ass, why don't you ask him how he come to this solution, where he is so confident. I mean after all, it should not be his opinion but based upon fact, as this is your money he is messing around with.

You have the chance to be a spokesmodel for the open-source community. I wish you well.

Last edited by rooloo (2008-08-15 10:57:31)

litemotiv · 2008-08-21 20:53:13

Update & FYI: The incompatibility problem has been acknowledged by OCZ tech support (seems like the other support guys mentioned above were 'mere' forum moderators who don't know what they're talking about). The drive is on it's way back to the supplier, and i will supposedly receive a full refund.

So here's a last final warning for people considering to buy a (relatively affordable) OCZ Core SSD drive: be sure it's compatible with your chipset/controller before getting one, at this point there is no guarantee whatsoever that it will be, in which case you will lose data. You might want to wait for the upcoming Core V2 series (firmware upgradeable) or the newly announced Intel MLC drives which should arrive shortly.

KimTjik · 2008-08-21 22:37:40

The Intel one might be expensive, but I suppose it at least means that it will be compatible and according to standard.

Anyway, thanks litemotiv for making the effort to get to the bottom with this issue. It'll for sure benefit others.

Arch Linux

#1 2008-08-11 21:15:35

Lost all data on home partition, would really appreciate advice

#2 2008-08-11 21:48:40

Re: Lost all data on home partition, would really appreciate advice

#3 2008-08-12 12:20:19

Re: Lost all data on home partition, would really appreciate advice

#4 2008-08-12 12:29:22

Re: Lost all data on home partition, would really appreciate advice

#5 2008-08-12 12:54:04

Re: Lost all data on home partition, would really appreciate advice

#6 2008-08-12 13:36:40

Re: Lost all data on home partition, would really appreciate advice

#7 2008-08-12 13:47:23

Re: Lost all data on home partition, would really appreciate advice

#8 2008-08-12 14:05:06

Re: Lost all data on home partition, would really appreciate advice

#9 2008-08-13 01:43:52

Re: Lost all data on home partition, would really appreciate advice

#10 2008-08-13 10:44:15

Re: Lost all data on home partition, would really appreciate advice

#11 2008-08-13 18:22:27

Re: Lost all data on home partition, would really appreciate advice

#12 2008-08-13 21:49:30

Re: Lost all data on home partition, would really appreciate advice

#13 2008-08-14 13:38:48

Re: Lost all data on home partition, would really appreciate advice

#14 2008-08-14 18:27:01

Re: Lost all data on home partition, would really appreciate advice

#15 2008-08-14 19:25:12

Re: Lost all data on home partition, would really appreciate advice

#16 2008-08-14 19:39:50

Re: Lost all data on home partition, would really appreciate advice

#17 2008-08-14 20:00:11

Re: Lost all data on home partition, would really appreciate advice

#18 2008-08-14 21:15:49

Re: Lost all data on home partition, would really appreciate advice

#19 2008-08-15 01:01:16

Re: Lost all data on home partition, would really appreciate advice

#20 2008-08-15 10:52:15

Re: Lost all data on home partition, would really appreciate advice

#21 2008-08-21 20:53:13

Re: Lost all data on home partition, would really appreciate advice

#22 2008-08-21 22:37:40

Re: Lost all data on home partition, would really appreciate advice

Board footer