This is not a hoax or a joke. There is apparently at ;east one trigger for a catastrophic event in Linux, which I stumbled upon by chance. A bad chance as it happened, as every effort to offset what followed failed miserably. I posted the history of my efforts, and the initial casue as file 100.txt at https://my.pcloud.com/#page=filemanager&folder=63120803
The mystery to me is what possible setting in a user account could act to bring a qhole system down by gradually impairing one application after another? One user account. Ubuntu only allows two at most, and the 2nd one continues to work well -- at least appears to. But who really knows what is happening inside?
Here is how to duplicate the problem:
(1) Set up a partition just for this test. Everything on that partition will be at risk, so make it just for this purpose.
(2) I installed Ubuntu 14.04, but if I am right, this could be an version of Linux. Some testing with different versions and distros should prove or disprove this.
(3) Install gnome-session-fallback, aka as Gnome Session Manager under the Ubuntu Software Center That's actually the name of the project. For Ubuntu 15.10, you have to settle for a subset called gnome-session-flashback. Of the three interfaces to use, pick Metacity, which was my choice. Or you could see it it happens with the other two interfaces as well. The impacts under Metacity affects them as well, and they may do the same thing on their own.
(4) Metacity works well, as far as it's performance is concerned. Some File Manager options under Full were removes in Flash, but you can use dconf-editor or gsettings to get things as you want them to be.
(5) To begin bringing it down, use you ability to change Metacity by going to either the top panel(bar) or the bottom panel(bar) and use Alt+Right MouseTouchpad Button. Select Properties. Ypu can do things like hide, autohide, buttons, arrows on buttons, or change colors and no problems. But to trigger the crash, you either move the bottom panel up, or the top panel down. I never tried to add a panel, so I don't know if that does the same or not/
That's it. You will get a crash and notice a report was sent off. You see things as normal. The moved panel will show up where you expected it to go. But the damage has been done, and over hours or a few days, your Linux install will begin to leak at the seams as one package after another quits working as it should.
And as I said, this only seems to effect the one user account, but I can't say that with any real certainty. I've been too busy fighting creaping disaster. It almost ended today when rm deleted everything on the partition instead of the one folder and file type I was after. That is when I learned of testdisk and tried it. I'm getting back far more than I wanted, but at least I am getting it back. Screening and filtering will come later.
What hurts though is the made up names, though I understand the issue there. But the need to keep changing parritions to catch what's coming seems only to show the current position, not other partitions or external drives. So I do that in another terminal window by moving recovered partitions off the current partition to free up space to recover more. Thakes longer that way, and it looks like I lost a number of recovered folders when space ran low on the current partition. But a lot recovered is better than a little lost.
Except I lost more again in my mv effort to get folders off onto a USB drive so there was space again on the current, meaning root, drive. The only real choice seems to be the folders under the user account on the user account, but that is still just the one drive. If I could get to the root / level, I can access other drive options by their mount points. How about a networked drive option as well?
The biggest help would be to trim the recovery of files by some filtering. There us no date association with when a file or folder was deleted, so I can't specify to disregard everything deleted more than a week ago, but I might like to set limits on files sizes as an example. Like all files over 20 GB are to either be ignored or recovered separately, or that any files of identical sizes over a certain size are to be compared to see if they are clones or not. I may only want image files og files that fit in the text category. I myself would be less interested in exe files as I'm using Linux. And as most bin files would likely be system files, these would be provided in the next install and not of interest to me per se.
I'm not complaining, mind you, My situation is such that debugfs would not even open the partition. So testdisk is going an incredible job. I;m going to try it next on a partition that grub screwed up when it attempted to resize a partiton to make room for an new Linux install. It left a file corrupted, and as it was a NTFS partition, there were no tools in Linux to fix it except, as far as I knew at the time, gparted. gparted spent 20 minutes trying to get past the trouble spot, and I fave up and had to install Windows under Virtualbox just to use chkdsk, the only tool it provided. But chkdsk was out of its depth, and I went back to gparted that seemed to be making progress. After most of a day, gparted resumed its scanning. But 9 days later, it still showed no times finishing up. And that was just a check, not a recovery of data. Now there may be some hope.
But you could take it further. Rebuild the partition as you go. Ypu can't get everything, and the writes to disk could be over what has already been dealt with. That way there is no need for lots of additional storage space to hold what is recovered. Now if a drive is going bad, or the user wants the option to keep the partition as-is for other recovery options, then you would need the added off-partition storage. But in my case, a write-back would have been a better choice. That's where all myext2.3.stuff is going to end up anyway. The partition could be ext2/3/4 for linux and one of Windows choices, though one is limited on size and the other is rather proprietary. However, there are toolss for Windows where they can at least read ext2/3/4 partitions. so maybe read-only under Windows? Though tools like gparted offer format conversions, and you can even use gparted from its own ISO or off many Linux LiveCDs.
Using TestDisk to repair the filesystem
1 post • Page 1 of 1
Who is online
Users browsing this forum: No registered users and 3 guests