Le 27.07.2014 01:42, PaulNM a écrit :
On 07/25/2014 10:54 AM, berenger.mo...@neutralite.org wrote:
First time I have exhausted inodes, but I never used apt-cacher-ng
previously, and it's quite obvious that a proxy+cache is very greedy
in
terms of inodes.
Not really. That's like saying the parking lot is greedy in terms of
parking spots, after you just drove a bunch of cars into it. :)
Indeed. I took a stupid shortcut.
Inodes are files/folders, files/folders are inodes. (1-to-1)
Anything
that has a bunch of files/folders will use a bunch of inodes. Same
number in fact.
Hum... is it accurate?
Files can use more than one inode, with ln. Folders can not, AFAIK,
since symlinks are simply pointers to inodes (which are themselves
pointers --with reference counter I guess, std::shared_ptr in c++11?--
to data).
I'm simply asking, I might be completely wrong or inaccurate...
The nice thing here is that I have learned a lot of this error, and
maybe someday I'll be able to help someone else in a similar
situation,
or be able to understand better partition systems.
Learning is good, keep it up. :) Others have already told you the
long
term fix (copy data, reformat, copy back), but there's another
option.
Inodes are a per-filesystem instance thing. If you can free at least
1
inode on /var, then you can:
create a file
mkfs.ext4 (or whatever) it,
temporarily loopback mount it somewhere,
move a large folder's (inode-wise) contents into it,
umount it, add it to fstab, then remount.
A bit complicated, but it's something you can do on the live system
without external drives. Technically the loopback mounted file
doesn't
need to be inside /var, but you have plenty of storage space there,
so
why not use it.
Very nice way! I did not know that it was possible to "insert" a
file-system into a file (but, finally, it does makes a lot of sense,
that's what iso files are finally).
I'll go for that solution.
I won't name it complex, it's simply not a classic procedure. Doing
things differently rarely stops me.
One of my defects is that I always try to tweak things... (with time
I've learned to not do that when the target is very important) but
at
least it allows me to learn. By failures :)
Yeah. Choosing the bigfile option when formating doesn't really save
drive space. It does simplify the filesystem records a little as
ther
are fewer records to keep track of.
In fact it can possibly end up using more space if you have a bunch
of
smaller files.
Yes, this is one of the few things I know about filesystems: I guess
that those options change the cluster width, which are the real space
unit when trying to create files. Now, I would never have guessed that
it also affect the number of inodes...
If you do reformat /var, I wouldn't use xfs. As others have
mentioned,
it has a few oddities that can cause issues if you're not fully
prepared. By all means create a sparsefile or regular loopback mount
to
play around with it, but for important stuff on your system, stick
with
what you know.
This would mean sticking with FAT32 hehe (since when I was a pure
newbie to computing, I played a little with winhex --and other reverse
engineering tools-- to open lot of stuff, including whole hard disks.
This tool is very nice, and I which I could buy it and use it on my
current favorite OS. For now I'm using wxhexeditor, quite nice, but the
author knows that he can't compete with winhex's devs.).
Using Linux is a way for me to learn things, and relying on Debian is
my way to have some trust that not everything will self-destroy while
I'm learning to do new things. Or at least that I will be able to repair
things I've broken, this being my main argument to not use windows
anymore.
But I'll accept your advice, using ext systems for important data, and
probably dig into other filesystems.
Taking the default values because I do not know what other means ashame
me, why would I do that on a system which owe it's strength to choice
and alternatives? (even the kernel can now be changed in Debian :) so I
guess we can no longer say that Debian is a Linux distro)
--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive:
https://lists.debian.org/80d3bca6e9a775d79e8721c052942...@neutralite.org