[NTLUG:Discuss] "Undo" hard links?

[Cornelius Keck]

Lance,

happy to know that this did the trick. :)

Unfortunately I have not much experience with external (as in 
off-premise) storage. Sure, my ChromeBook puts a very few things there 
(like my first attempt at https://dallasrpi.org/ created using Google's 
tools to create a simple website, but the generated code not only was a 
bit too chunky for my needs, it also contained all kinds of links to 
Google's unwanted advertising and/or tracking, so I decided to use 
SeaMonkey's Composer instead, which tells you all about my web design 
skills), but that's pretty much it. My phone keeps nagging that I should 
set up iCloud.

It doesn't help that I don't trust outsiders having access to my data, 
and I've been seriously debating putting drives into a bank's vault. Not 
exactly online, but somewhat safe from intrusion. Behold, the return of 
sneaker net.

Depending on how much you like your workplace, keeping a copy there 
might do the trick. 4GB is not enough to attract unwanted attention, and 
there might be places that are not under admin surveillance.

Setting up a public-facing server isn't as tricky as it sounds, dynamic 
DNS probably the tricky part. Rest is straightforward, with a bit of due 
diligence. I self-host on static IPs, now for 20 years, and while there 
have been more than enough attempts to get in here, they have yet to 
succeed. Downside is, for geographic diversity you need real estate.

Everybody else on here, chime in, ideas please :)

-Cornelius


Lance Simmons wrote:
> Cornelius, thank you  very much! I've been so busy I didn't get to try 
> this, but it worked just fine.
> 
> I already do local backups on two different internal drives and one 
> external drive. I use Google Drive so that I can also have multiple 
> backups at work. (I do almost all my work at home.)  I also do 
> incremental backups at home, but that takes too much space for my free 
> Google account. At a bare minimum, I want backups of my documents in two 
> different physical locations, and at least one of them incremental. 
> (We're talking about a whopping 4 GB, but it's about 25 years of my life.)
> 
> Is there a good alternative to Google Drive? Gnome has support for 
> Nextcloud built in, but a cursory look for Nextcloud providers didn't 
> find any cheap hosting.  I'd rather not spend the time and effort to set 
> up dynamic dns and have an outward-facing server.
> 
> Thank you again! This was exactly what I was looking for.
> 
> On Tue, Nov 10, 2020 at 6:29 PM Cornelius Keck <dfwuug at keck.us 
> <mailto:dfwuug at keck.us>> wrote:
> 
>     Yes... brute force might do the trick. Make sure you have enough disk
>     space. Given this scenario:
> 
>     pi at pi1b00:~/tmp $ ls -ls
>     total 12
>     4 -rwxr-xr-x 1 pi pi  139 Nov 10 18:12 breakLinks.sh
>     4 drwxr-xr-x 2 pi pi 4096 Nov 10 18:11 D1
>     4 drwxr-xr-x 2 pi pi 4096 Nov 10 18:11 D2
>     pi at pi1b00:~/tmp $ ls -ls D*
>     D1:
>     total 0
>     0 -rw-r--r-- 2 pi pi 0 Nov 10 18:11 f1
> 
>     D2:
>     total 0
>     0 -rw-r--r-- 2 pi pi 0 Nov 10 18:11 f2
>     pi at pi1b00:~/tmp $
> 
>     of two subdirectories, D1 containing one file, D2 a hard link to that
>     same file (link count of f2 is the same as the one of f1, and larger
>     than 1 (one)), run this script:
> 
>     pi at pi1b00:~/tmp $ cat breakLinks.sh
>     #!/bin/sh
>     find . -type f -links +1 -print | while read line
>     do
>           cp "$line" "$line.tmp"
>           rm -f "$line"
>           mv "$line.tmp" "$line"
>     done
>     pi at pi1b00:~/tmp $
> 
>     This looks for files (type f) in the current directory, with more than
>     one reference, i.e. inode reference count larger than 1 (one) (-links
>     +1), then, for each found, copies it aside, removes the original
>     (reducing the inode's link count), moving the temporary copy back into
>     place. The quotes around $line are there to keep the script from acting
>     up when running into file names containing spaces.
> 
>     Running that results in:
> 
>     pi at pi1b00:~/tmp $ ./breakLinks.sh
>     pi at pi1b00:~/tmp $ ls -ls D*
>     D1:
>     total 0
>     0 -rw-r--r-- 1 pi pi 0 Nov 10 18:23 f1
> 
>     D2:
>     total 0
>     0 -rw-r--r-- 1 pi pi 0 Nov 10 18:23 f2
>     pi at pi1b00:~/tmp $
> 
>     Same as before, but the link count is 1 (one).
> 
>     If the files' timestamp is important, use "cp -p".
> 
>     Sure, one could do this in a one-liner using find's -exec option, but
>     that's makes for a less readable explanation.
> 
>     On a personal note, Google Drive is hosted on somebody else's computer,
>     so there is always the chance that somebody else might get to your
>     data,
>     or that Google turns off access on short notice, or changes its terms
>     and conditions. While it is a convenient feature, from a privacy and
>     security perspective, a local backup is preferable. Besides, external
>     disk drives are fairly cheap, Black Friday is coming up, and chances
>     are
>     that your uplink is not as fast as a USB3 wire.
> 
> 
>     Lance Simmons wrote:
>      > I recently created a lot of hard links, thereby reducing the disk
>     space
>      > used by my Documents directory by a chunk. I was feeling pretty
>     good about
>      > this, but then discovered that Google Drive does not acknowledge hard
>      > links. Every time the folder is synchronized (via a cron job), a few
>      > hundred bogus "conflicts" are declared and  then resolved. Then,
>     the next
>      > time the folder is synchronized, the same "conflicts" are found and
>      > resolved again.
>      >
>      > I love the idea of hard links, but now I want to return to having
>     separate
>      > files. Is there an easy way to (1) delete hard links, then (2)
>     copy the
>      > files (inodes), giving them the filesystem names of the original hard
>      > links?
>      >
>      > I should switch away from Google Drive; maybe this annoyance will
>     motivate
>      > me.
>      >
>      > Lance
>      >
> 
> 
> 
> -- 
> Lance Simmons