57 thoughts on “Strange Computer Problem”

1. Power cycle? Maybe something went south in the SSD firmware.

   1. Or the hard drive firmware more likely.

      1. You might also have a marginal hard drive. Recommend immediate back up just in case. That could take forever unless you have it shadowed somewhere. If the power cycle doesn’t help or makes it worse time to consider a swap out of /dev/sdX that hosts /home….

         The reason I focus on /home is that if you had done a complete reboot the modern ext3 & ext4 file systems do journaling. You should not be experiencing fsck issues across a reboot under those file systems unless there is some underlying hardware issue. Assuming a clean reboot that you didn’t interrupt…

         1. That does seem to be the problem. I hope I can back it up.

   2. I did shut down and restart.

      1. Shutdown doesn’t imply power cycle. Did you power cycle?

         1. In Linux lingo a shutdown merely means the kernel has entered a halt loop. On older timesharing systems a shutdown implied dropping into single user mode as root with the kernel still running. A shutdown with restart means some piece of software in the kernel causes an unmaskable CPU interrupt to trigger a reboot. None of these scenarios cycle power to either the CPU or motherboard leaving asynchronous state machines and / or bus driven microcontrollers to their own mystical ways and affectations.

2. How was the speed when you booted from stick?

   1. It was fine.

      1. Assuming a power cycle see comment about disk hosting /home

3. A guide for doing disk forensics under Linux. HTH

   https://fedoraproject.org/wiki/Smartctl

   1. Using that tool, the disk self reported that it was fine. But if I mount it at boot, the machine doesn’t want to let me log in…

      1. smartctl is far from perfect it’s biggest usefulness is to watch the reported errors to see if that is incrementing frequently. Sometimes you can get a result by engaging the long test and then see if the error log shows bigger numbers than it had before. Usually the self test won’t fail you have to watch the behavior of the error log. It’s far from foolproof but is one way of conducting a non-destructive test. Regrettably with Linux you have to look for symptoms. It won’t tell you outright.

         1. Well, now the disk management tool in Gnome is telling me that the disk is self reporting as failing. I have another drive I want to back it up to, but it has two partitions on it: a boot and the OS. I deleted the boot partition, but the rest is still calling itself /dev/sd2, when I want it to be /dev/sd1. Any way to fix this, short of blowing away the other partition with all its existing data (it’s an old backup)? If I can’t renumber it, the boot will fail when it tries to mount it.

            1. See:
               https://superuser.com/questions/393613/how-to-renumber-a-partition#915412

               You can renumber the partition table by entering the ‘x’ command to enter ‘expert’ mode.

               Before you do the ‘w’ make sure you’ve written down or copied all the info from the first -l so you can put the partition table back to something useful if you mess it up.

               I haven’t done this in a few years so I’m reticent to give you too much advice here. Lest I get it wrong.

         2. The smartctl self test is driven by the disk vendor’s firmware. And frankly it appears most vendors can’t be bothered with a decent self test if they bother at all. Something even guides like Tom’s Hardware ought to make an issue of but hey $/bit is Uber alles!

            1. I see that Best Buy has a Western Digital 2 Terabyte for fifty bucks, so I’ll just back up to a new drive.

            2. +42

            3. I found a vintage IBM PC 70 MB full-height drive on Ebay for $8.50. Proven, reliable, and offering plenty of storage space for documents, recipes, and games. But they want $14+ for shipping, and I’m not sure your PC can mount a full-height drive. So you’re better off with the 2 Terabyte one.

            4. Sometimes we just have to make due with less.

4. Try dd again with the arguments used for disk to disk in this example:

   https://www.thegeekdiary.com/how-to-backup-linux-os-using-dd-command/

5. You have more problems with Linux in a year than I do with Windows in a decade.

   1. To be fair this is a hardware issue.

   2. This would have happened regardless of the OS.

6. <I.I got the new drive, and started to dd the data from the old drive to it. The process died after about 2.7G, with an “i/o error.” How screwed am I?

   You’re “this is why we make backups” screwed, at least potentially.

   But it’s a useful lesson – if it ain’t backed up, it’s already gone is the one true way to approach data.

   1. True, perhaps an investment in a network drive might be in order?

      1. Or invest in another WD disk and set the two up in a RAID array with one being the shadow disk. Either way.

7. Well, it’s copying at 11MB/s. At that rate, it’s about a third of the way through, and won’t be done until tomorrow. I’m glad it wasn’t bigger…

   Assuming you are copying from rotational media this sustained slow rate is also telling. Rather than having a bad block or sector sounds like a sense amplifier is having great difficulty. Which makes multiple accesses necessary to pull the data off the recalcitrant disk which means you can only pull data off at a fractional rate of your I/O bus whether SATA or something else. Good luck but most of all be patient!

8. After a painful experience with a failed drive, I set up an old refurbished computer to continually back up important parts of my home directory. I set up network drives and use fwbackup (a front for rsync). Works great.

   I also like to always have a relatively up to date PartedMagic drive (with Clonezilla and Ghost 4 Linux) for cloning drives. Clonzilla makes more compact images, but it’s more fussy. Given your disk errors, the more brute force G4L would likely be more appropriate. It’s like a front end for dd, but it’s easy to use.

   Finally, it makes sense to have a copy of Boot Repair on hand, but you know that.

   1. +42

9. Try the ddrescue tool if the dd copy doesn’t work out. It has a more aggressive retry policy and also logs the status of each copied region.

10. Well, generally a drive is partitioned and then formatted before it is useable, the old “FORMAT C:” solution to many a software issue. ^_^

11. It is distinctly possible that I didn’t format the drive after creating the partition. I’ll try again in the morning.

    1. It’s events like these that make me happy I’m on the East Coast and fast asleep…. *_*

12. “Yes, I forgot to format after petitioning…”

    You have to “petition” your computer?

    Um, I think you would be much better off with Windows 10, or even with some incarnation of an Apple OS – Jaguar. Leopard, or even Ferrari, and, especially, Blue Steel.

    1. When you can pry my Linux from my cold, dead, hands…. 🙂

       FYI Apple OS *is* essentially Linux, under the hood….
       Drop into the command line interpreter and poke around if you need proof… like % cat /proc/cpuinfo for example…

       1. Been that way since System X (System 10) I believe…
          I once had the privilege of programming directly on System 6. It was fun and educational to see the lineage of Andy Hertzfeld’s work. Nicely done Andy…

       2. Or even better: % uname -a
          heh…

          1. Erm I mean: GNU/Linux sorry rms….

13. Um save the drive that had the 833GB transfer just in case. Your source disk sounds like it is really really on its way out. Did you get enough via the first transfer that all key files were recovered? Just sayin’ this might be the best you can do. Highly recommend a backup strategy. And of course my Space Cadet training says that I have to ask the following question: If you have one, why not just restore from it? Not trying for the salt->wound, just trying to save you time…. There’s a little bit of the Doc McCoy in me that says a little pain is good for the soul….

    1. cp -r to selectively restore what is most important might be your friend here, just sayin….

14. And why is there “no space left on device”?
    Did you mount the output device? The path name you gave for the output file is suspicious. Usually when you mount a drive you have to give it a mount point in /mnt. Like /mnt/newdrive/mydiectory/test.img

    But why back up to an image file? Don’t you just want to go disk to disk? If you are doing direct can’t you just give it /dev/sda1 (or wherever you new drive configured) for the output file instead?

    1. $ info ddrescue.

    2. I was following directions here. Yes, I mounted a spanking new partitioned and formatted drive to /mnt. That’s where the *.img file went.

       1. That’s right, I forgot, you flipped on the benefit of coffee again.
          Well maybe you are caffeine immune? In that case, you live in CA, go outside (everyday is warm and sunny in CA right?) run around the house three times (in both directions for a total of six) THEN proceed with this…. ^_^

          I was following directions here.

          Jump ahead in those instructions to the section called “Cloning directly to a new disk”. This is what you need to do. Creating a dot img file is fine for when you are just doing backups, but you need a backup AND restore. To paraphrase DEVO: “backup is what you’ll get but cloning is what you need. The disk you want”. In fact playing DEVO in the background while doing this seems quite appropriate. (I know, I know you despise DEVO, ’cause of its Kent State genesis and all that…).

15. See this link for disk to disk recovery. But first pour yourself a big cup of Joe:

    https://datarecovery.com/rd/how-to-clone-hard-disks-with-ddrescue/

    1. I hate coffee, and it has no discernible affect on me.

16. ddrescue copies the entire device, so the destination device must be equal to or larger than the source. You mention you’re using less than 1 TB, but that’s a filesystem thing. ddrescue wants to copy the entire 1.8 TB verbatim. Like dd, it is copying sectors, not files.

    1. Yup also put the log file someplace else like /root.

17. A suggestion: In future posts about computer problems (and there WILL be such posts) it would make your blog easier to follow if you put the output of troublesome console sessions down here in the comments and just stick a one paragraph summary in the OP with a reference to comments for details.

    1. I’ve added a “more” tag.

    2. I meant to add that bumped short one paragraph summaries in the OP are fine until they get to multiple pages when you can then more tag them. It’s nice to be able to follow your progress. Good Luck!

18. So…. How’d you make out? All fixed?

    1. It’s still working on it. About 77% recovered, with estimates to completion ranging from a couple hours to a month, depending on current speed… Once I get it off the bad drive, I’ll probably back it up, and then do a file-system repair.

       1. I have a pair of 4 or 5 TB external USB drives which I should really make more periodic use of for full system backups. They’re cheap and if I traveled much they’d come in really handy for keeping all my data at hand.

          1. Since he now has two drives of the same size he can shadow one in an internal RAID array to leverage the speed of his I/O bus. Switchover ought to be painless and no restore needed although both drives age at the same rate. Nice to occasionally swap them each out with a newer drive to stagger their age. Like drives like tires. heh. Or go all SSD. The aging characteristics / data look good so far. If you can afford them. Age staging still a good idea even with SSDs that have no moving parts.

19. Pingback: More Computer Fun | Transterrestrial Musings

Comments are closed.