So my SSD just died after I dropped my laptop several inches. My last backup was 6 months ago.
The worst part was when I restored my system image (from the SSD to a different hard drive), my hard drive became G: and had the wrong drive letter, and Windows XP wouldn't boot anymore. Took a lot of work to get it to boot again.
What worked:
* Using Ultimate Boot CD For Windows:
* mount the C: drive using Microsoft DiskPart (Disk Management wouldn't mount it for some reason)
* Loading the System registry hive (C:\windows\system32\config\system), and copying registry keys from HKEY_LOCAL_MACHINE\System\MountedDevices to the real registry hive
* Using sfdisk on linux to make sure the partition ID was NOT set to 82 (linux default). Damn you gparted!
So BACKUP YOUR DAMN FILES. No excuses. Buy bigger hard drives if you don't have anywhere else to put them.
I use hard drive images as my form of backup.
Can you describe what observations you've made that lead you to conclude that the SSD died? Also, when you probe it from a working machine, what does it see? No device at all, one that doesn't allow any reading, etc.?
First, there was a BSOD about 20 seconds after the laptop hit the table, presumably it wasn't trying to access the disk before then. Then the ThinkPad BIOS said there was a hard drive read error.
I then tried plugging it into two low-end Intel motherboard machines (Atom machine and Celeron machine). They wouldn't even let me into BIOS setup with the drive plugged in. One just crashed, one gave a "No boot device" error message (different from the one you get when the BIOS finishes running and there's no boot device at that point).
I tried hotplugging it (probably a mistake). Then it became an 8MB drive that only reads zeroes, but then the BIOS setup ran without crashing or giving a "no boot device" error message.
It was an Intel X25-M 80GB SSD.
Can't explain earlier behaviour, but the 8MB capacity bug is known (search Google for "
Intel 320 8MB") and was fixed with a firmware update. I'm aware you have an X25-M, but given that the 320 came out *after* the X25-M, I'm inclined to believe the earlier models could suffer from the same behaviour. The bug got tickled when you hotplugged it (users experiencing the 320-series 8MB bug often did not disclose the fact that they were shutting their systems down abruptly (yanking power, hard power-off, etc.)).
Dwedit wrote:
So BACKUP YOUR DAMN FILES. No excuses. Buy bigger hard drives if you don't have anywhere else to put them.
I use hard drive images as my form of backup.
Thanks for the reminder, I've been putting off creating a personal backup system for the last... uhm... year or so. Maybe I'll finally get around to it. As it is I only have my most important projects under version control on a remote server (using Git).
Part of a backup strategy is to separate the files that you can easily replace from those that you've created yourself. Generally the latter category takes up a fraction of the space as the former, unless you create music or videos. Only if you have a huge backup drive can you just blindly backup everything.
blargg wrote:
Part of a backup strategy is to separate the files that you can easily replace from those that you've created yourself. Generally the latter category takes up a fraction of the space as the former, unless you create music or videos. Only if you have a huge backup drive can you just blindly backup everything.
I find Dropbox to be a great FREE solution for these file. Not only do I have nearly instant back ups online with some version control, but I've also got backups on every pc I've got Dropbox installed on. It's carefree and I don't have to do anything special to store/access the files.
blargg wrote:
Part of a backup strategy is to separate the files that you can easily replace from those that you've created yourself. Generally the latter category takes up a fraction of the space as the former, unless you create music or videos.
I created the video of your flowing palette demo, and I've created several other NES-related videos as well. A 2 GB Dropbox might not be quite enough to hold my entire oeuvre. But now that the market has corrected itself a year after historic flooding in Thailand, I'd bet most people can fit what they have created into the terabyte of space in a $100 USB hard drive.
I put my stuff to google drive every once in a while. That's about it. Plus I have 2 partitions on my HDD with the same info, and then my flash drive with backups too.
And the lesson here?
1. Backup your stuff as often as possible.
2. Be careful when handling laptops as you can never 'drop' your PC unless during a relatively rare move.
3. SSDs (if that is what is at fault here) aren't as safe as people think.
Most of my personal projects I put on an free private subversion repository at
http://assembla.com. Public projects are usually on Google Code or something. That takes care of my most important backup needs, day-to-day.
Stuff that doesn't fit these categories (e.g. long term storage, personal media collection, etc.) I manually copy to a pair of raid 0 hard drives that are in a relatively safe location. This comes up much less frequently. I'm considering hiring some sort of online storage service as a secondary backup for those things, but the cost and initial bandwidth required to do so has so far kept me from getting around to it.
Yesterday I made a full backup of my work drive, prior to doing a full reinstall. My last backup was from march... Got to do them more often !
TmEE wrote:
Yesterday I made a full backup of my work drive, prior to doing a full reinstall. My last backup was from march... Got to do them more often !
At least something good is now coming from somone else's misfortune.
I did not know about his problem, I moved my PC inners to a new chassis and I thought I'd do a reinstall while I was at it to get 30 second boot again instead of 2 mins...
TmEE wrote:
I did not know about his problem, I moved my PC inners to a new chassis and I thought I'd do a reinstall while I was at it to get 30 second boot again instead of 2 mins...
Once you go SSD you never go back.
I have really bad luck with hard drives (especially WD), so I just use RAID mirroring and be done with it. It seems expensive at first (double the price for the same capacity), but when you hear a drive going clickity click, you know you don't have to waste a day setting up your environment again. I use the Intel Matrix RAID because the drive is still readable if you plug it into any old SATA controller. The RAID data is also on the drive (not on the controller NVRAM), so if your controller bites it you're not SOL.
I still use Dropbox/Google Drive on my really important stuff (code, projects, etc).
But be careful what kind of RAID you use. Consider what happens when a drive fails, and another drive fails before the array finishes rebuilding. In RAID 5, you lose everything and have to go back to tape or the cloud. In RAID 10, you have only a 33 percent chance of losing everything: only a 1+2 or 3+4 failure causes loss, while a 1+3, 1+4, 2+3, or 2+4 loss is survivable. RAID 5 makes me wanna
BAARF.
Having used RAID-1 for the past 7 years, all I can say is that RAID has never saved my data.
What RAID does is keep your machine on and running after a hard drive failure. But I've lost data to hard drive failures once in 25 years, ever, and it would have been solved by real backups instead. And meanwhile I've lost data ~30 times to my stupidity^W^W and at least 8 times due to filesystem corruption.
The infamous JWZ says it more bitterly/amusingly than I would:
http://www.jwz.org/blog/?p=801607
I've had 3 hard drive failures, myself.
But yes, having your data in two locations is a lot safer than having two copies of your data in one location. RAID redundancy is no substitute for backups, but it's still good protection against failure of a single drive (which is a common enough occurrence that I think it's worth doing in some cases).
I like JWZ's article. Unlike everything else in your house, when your data gets lost, you can't buy replacement anywhere. You know the frustration when you've typed a long post and your web browser eats it before you click Post? Imagine that multiplied by 100000, all because you didn't take the time to make a backup (write your web post in a text editor rather than the browser).
tepples wrote:
RAID 5 makes me wanna
BAARF.
I think that RAID6 should generally be favored over RAID5, but saying that parity-based RAID levels should never be used (or, according to one of the presentations from that site, that RAID10 should ALWAYS be used) doesn't make sense. In the real world, one usually needs to make trade-offs between cost/performance/reliability. If I need a large amount of storage for an application with low IO requirements, RAID10 is probably not the best choice. Certainly not from a cost perspective.
blargg wrote:
write your web post in a text editor rather than the browser
And I do that for several reasons:
- To protect against web browser data loss, as text editors are usually simpler and less complex.
- Starting a text editor is sometimes faster than starting a web browser. Some sites are so slow that by the time the post form has opened, I've already hammered out half a paragraph at 80 wpm. Or I might have just opened my laptop, and it takes 30 to 60 seconds after coming out of sleep to find a wireless access point.
- I commute to and from work on the bus, and I don't feel like paying half a thousand dollars per year for mobile Internet on top of what I already pay Comcast for Internet at home. Often I start writing while offline.
But just starting composition in a text editor doesn't protect against the power going out or a kernel or window system failure, seeing as most PC text editors don't automatically save documents that haven't been given a filename.
James wrote:
If I need a large amount of storage for an application with low IO requirements, RAID10 is probably not the best choice. Certainly not from a cost perspective.
I guess part of their cost calculation includes the cost of downtime while restoring from slow backup media.
tepples wrote:
James wrote:
If I need a large amount of storage for an application with low IO requirements, RAID10 is probably not the best choice. Certainly not from a cost perspective.
I guess part of their cost calculation includes the cost of downtime while restoring from slow backup media.
Perhaps. Though, arguably, if downtime is that costly, you're looking at an HA solution (e.g., a duplicate system). The point I was trying to make is that there is no one-size-fits-all solution here.
Well, RAID-1 has been working fine for me for years. Drive fails, swap in another one, keep on going... If I were to loose both drives, I wouldn't have lost anything that can't be replaced, since all _MY_ files are on Dropbox and Google Drive. What RAID-1 saves me is all the trouble of reinstalling the OS and setting up my environment.
I'm aware you can rsync on Mac/BSD/Linux because the file system doesn't lock files like Windows. Unfortunately, I'm using Windows.
I prefer to think of it this way:
If your house burnt down while you were away, would all _YOUR_ critical data still be backed up?
95% of the data on my drive is just Music/Games/Movies that take up lots of space and can easily be replaced. Yeah, its a pain to download it all again, but I haven't lost anything permanently.
RAID protects from some hardware failures, downtime. But it's an exact mirror, so doesn't save you from non-hardware-caused data loss. Separate backup protects you from hardware failure, and from accidentally deleting files, filesystem corruption, etc. Off-site protects from theft, house fire, etc. Also, incremental backups protect from undetected corruption that gets copied to the more recent backup(s).
Dwedit how far did you drop it and since it has no mechanical parts what would cause it to break?
WedNESday wrote:
Dwedit how far did you drop it and since it has no mechanical parts what would cause it to break?
No offence intended, but at this point it doesn't matter -- the SSD has now experienced the 8MB capacity bug and the data stored on the drive is lost.
It's possible Dwedit could give the SSD to one of the very few data recovery companies that restore data from SSDs (specifically they give you raw dumps of each NAND flash chip), but it's expensive (more so than MHDD recovery), and if the drive internally uses encryption (often AES) or compression for its storage there's still no chance of getting any of the data back since raw NAND flash dumps don't provide FTL mappings (i.e. LBA-to-NAND-page mappings). The X25-M is somewhat "old" and AFAIR does not use encryption, but like every SSD, does rely on an FTL.
Moral of the story: do not think even for a minute that SSDs are impervious to physical damage. They aren't. Just because something lacks moving parts doesn't mean something can't be broken when given a good jostle.
Do regular backups. And like others have already said: RAID is not a replacement (nor a supplement) for backups.
does it the SSD have parts on both sides of the PCB?
If it only has parts on one side of the pcb and you've already considered it a loss, a fun last ditch effort might be to toss it in a toaster oven to reflow the whole board.
If nothing else you can just leave it in there for way to long over baking it while you tell it to, "BURN IN HELL!"
infiniteneslives wrote:
does it the SSD have parts on both sides of the PCB?
The X25-M
has parts on both sides -- the controller and some NAND flash chips are on one side, and more NAND flash chips are on the other. Both sides have miscellaneous parts (capacitors, resistors, etc.) as well. For SSDs this is normal (NAND flash going on both sides of the PCB).
My RMA replacement from Intel just arrived. I think they sent me a newer drive, it's a "320 series" instead of the X25-M, still 80GB.
In a rather thick fit of irony, I was doing an NTFS resize on my hard disk drive (120GB) to get it ready to ghost to the 80GB SSD drive, and the filesystem was corrupted after the NTFS shrink, even though NTFSResize said it was successful. Linux wouldn't mount it, and neither would Windows.
I restored the 80GB SSD from the exact same 6 month old backup, and this time Windows booted on the first try. It downloaded about 38 Windows Updates soon afterwards.
Dwedit wrote:
In a rather thick fit of irony, I was doing an NTFS resize on my hard disk drive (120GB) to get it ready to ghost to the 80GB SSD drive, and the filesystem was corrupted after the NTFS shrink, even though NTFSResize said it was successful. Linux wouldn't mount it, and neither would Windows.
I never trust any filesystem operation like that, even if it's claimed to be 100% safe. Same for defragmenters; stay away from my disk!
Dwedit wrote:
In a rather thick fit of irony, I was doing an NTFS resize on my hard disk drive (120GB) to get it ready to ghost to the 80GB SSD drive, and the filesystem was corrupted after the NTFS shrink, even though NTFSResize said it was successful. Linux wouldn't mount it, and neither would Windows.
This happened to my fiancee's laptop shortly after she got it, in the exact same way. Two sets of system restore disks later (originally: 7 professional 64-bit. first: 7 pro 32-bit. second: 7 ultimate 64-bit) she finally got a functioning windows install again.
blargg wrote:
Same for defragmenters; stay away from my disk!
I kinda like xfs_fsr for that reason.
Dwedit wrote:
My RMA replacement from Intel just arrived. I think they sent me a newer drive, it's a "320 series" instead of the X25-M, still 80GB.
In a rather thick fit of irony, I was doing an NTFS resize on my hard disk drive (120GB) to get it ready to ghost to the 80GB SSD drive, and the filesystem was corrupted after the NTFS shrink, even though NTFSResize said it was successful. Linux wouldn't mount it, and neither would Windows.
I restored the 80GB SSD from the exact same 6 month old backup, and this time Windows booted on the first try. It downloaded about 38 Windows Updates soon afterwards.
Norton Ghost does a very good job of re-sizing during a clone operation of NTFS images without pre-re-sizing the image before-hand, in my experience, if that could help in the future.