Moving/merging a filesystem back into /

Kevin Korb kmk at sanitarium.net
Tue Dec 3 08:17:06 MST 2013


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 12/03/13 10:05, Charles Marcus wrote:
> On 2013-12-03 9:37 AM, Kevin Korb <kmk at sanitarium.net> wrote:
>>> Or do I need to specify the filesystem type?
>>> 
>>> mount -t ext3 -o noatime /dev/sda3 /mnt/gentoo/ ?
>> I would expect to do this.  Also, you should be converting to
>> ext4 not ext3.  In fact, your kernel probably doesn't even have
>> ext3 in it anymore.  Standard kernels now use the ext4 code for
>> ext3 mounts.
> 
> One thing at a time... ;)

Sure.  Simply put, you can just change ext[23] to ext4 in fstab and it
will just work.  Faster even.

> It is working now, and currently has ext3 in the fstab. I'd rather
> not change that at the same time as doing something like this.
> 
>> Concerns no. Lack of terror, yes. You will no longer be exposed
>> to an fsck that is almost as destructive as a mkfs
> 
> Yes, I've read the horror stories. I've also had this system
> running for 8+ years, and had 2 hard resets (and successful fscks)
> during this time...
> 
> There is a lot of FUD about reiserfs.

Reiserfs can take a hard crash without needing an fsck just like
ext[34] can.  This is what journaling is for.  I am talking about when
fsck -p aborts and says you need to run fsck manually.  In ext[34]
this is a minor inconvenience and you can usually even run fsck with
- -y to just say yes to any prompts.  With reiserfs if you ever have a
problem at this level you are screwed.

> Not saying it is perfect, or that your concerns are misplaced (my
> new system will use xfs instead), but it isn't as bad as you're
> making out... ;)
> 
>> Nonsense. --numeric-ids disables the ownership translation
>> between 2 different systems when rsync is networking. When you
>> are booted from a live environment you still only have 1 user
>> database. It might disagree with the one you normally boot from
>> but you still have only 1 so there is nothing to translate.
> 
> Ok, well, like I said, since it doesn't matter, that is an argument
> I may pursue later, if I fell a need...
> 
>>> Doesn't matter though, since it can't hurt anything, I'll
>>> always use it regardless... ;)
>>> 
>>>> ext[34] can remember acl as a default mount option. I have
>>>> no idea if reiserfs can do that or not and I recommend
>>>> against using reiserfs at all.
>>> I know, but this system was built over 8 years ago (when
>>> gentoo still recommended reiserfs), and runs like a champ, so
>>> no desire to 'fix what ain't broke'... anyway, acls are not
>>> enabled on it, so no worries...
>> If you ever have an instance where reiserfs needs an fsck you may
>> as well reformat and restore.  Reiserfs's fsck just deletes
>> almost everything then declares success.  Reiserfs was never
>> recommended except for /usr/portage which can always be
>> re-downloaded.  It should absolutely never be used for anything
>> important.
> 
> Sorry Kevin, this is just plain BS.
> 
> When I installed gentoo, it was recommended for things like mail
> storage when you were using maildir (lots of small files)...
> 
> And as I said, my experience directly refutes your silly claim
> that "Reiserfs's fsck just deletes almost everything then declares
> success".

I have seen it happen.  It is rare that reiserfs needs an fsck but if
it does the fsck is destructive.

> It has had its problems, and indeed may under certain circumstances
> be more likely to experience unrecoverable damage to the
> filesystem, but most serious problems it has had were fixed fairly
> quickly (but maybe not after lots of people got bit). The fact is
> ext4 has had its problems and horror stories too, just as xfs and
> every other filesystem out there. It is best to stick to verifiable
> facts when discussing things like this.

I wouldn't recommend XFS on Gentoo either.  The XFS development is
mostly done by RedHat these days.  It tends to only work completely in
1 out of 5 or so kernel versions.  This is fine if you are on RedHat
because they make sure it works on every kernel version they release
but the same is not true on Gentoo.  I used to use XFS for my media
storage (it is great on large files) but I had to give it up because
it had a bunch of weird problems interacting with NFS.

>>> emerge --info shows neither:
>>> 
>>> USE="3dnow acl amd64 bash-completion berkdb bzip2 cli cracklib 
>>> crypt curl cxx dovecot-sasl dri fam fortran gd gdbm iconv mmx 
>>> modules mudflap multilib ncurses nls nptl openmp pam pcre
>>> readline sasl session snmp sse sse2 ssl tcpd truetype unicode
>>> vhosts xml zlib"
>> After you finish with this, try adding caps to that.  It is a 
>> significant security improvement IMO.
> 
> Any pointers to docs explaining how/why? I'll google later, but
> I'm curious...

http://www.friedhoff.org/posixfilecaps.html

>>>> Even on something more dynamic like /home you can still do
>>>> an initial copy then update it from the live environment.
>>> 
>>> Ok, so... if I wanted to do this, would I need to add anything
>>> to the rsync command on the subsequent run(s)?
>> nope.
>> 
>>> So, looks like the command I'll be using:
>>> 
>>> rsync -avHP --numeric-ids /mnt/gentoo/oldusr/ /mnt/gentoo/usr/
>> yep.
> 
> Cool, thanks again Kevin... :)/**//* */
> 
> 

- -- 
~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~
	Kevin Korb			Phone:    (407) 252-6853
	Systems Administrator		Internet:
	FutureQuest, Inc.		Kevin at FutureQuest.net  (work)
	Orlando, Florida		kmk at sanitarium.net (personal)
	Web page:			http://www.sanitarium.net/
	PGP public key available on web site.
~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iEYEARECAAYFAlKd9fIACgkQVKC1jlbQAQdOMACg2zvGakPDhvxFT71lqkEd4FXq
ATEAnj42zfy0IbDJYnRu3FMD01N/d/f6
=MlnS
-----END PGP SIGNATURE-----


More information about the rsync mailing list