From: "Theodore Y. Ts'o" <tytso@mit.edu>
To: "Rogério Brito" <rbrito@ime.usp.br>
Cc: linux-ext4@vger.kernel.org
Subject: Re: Generic questions about ext4
Date: Mon, 17 Aug 2020 15:43:44 -0400 [thread overview]
Message-ID: <20200817194344.GA6755@mit.edu> (raw)
In-Reply-To: <CAOtrxKPMpNh7MZ2wX2wxju15rrY9rzrjNAwFneyAwCy2Z7DVMQ@mail.gmail.com>
On Mon, Aug 17, 2020 at 03:54:29AM +0000, Rogério Brito wrote:
> Dear developers,
>
> I have been using ext4 for quite some time and I have some ext4
> filesystems that led me to some questions.
>
> I have at least 2 "large" filesystems (with 2TB each, both almost
> full) dating back from 2011 and 2010. I believe that I even converted
> one of them from ext3 to ext4, but my memory is not that clear after
> almost a decade. As such, they were created without some useful
> features that are useful nowadays, like inline files. With that in
> mind, here are some questions:
>
> 1 - I know that some features can be enabled with tune2fs, but, in
> particular, inline files don't seem to be. I've seen some people
> indicate that using debugfs, I can mark the superblock as having
> support for it. Would that really work? I don't plan on booting old
> kernels. One of those filesystems is running on an armel device that
> is quite slow and I would really like to avoid copying all the files
> to an external HD, recreating the filesystem and, then, copying back
> the files to the system.
Inline data is not enabled by default, because there are still a few
corner cases that will cause stress tests to fail. So it's not
something I'd encourage enabling; it doesn't make that much difference
unless your workload creates huge numbers of small (< ~100 bytes)
files. If you do create lots of small files, but then don't attempt
to append to them, or truncate them, etc., later., inline_data will
work, but it's likely that the presence or absence of inline_data
probably won't make that much difference to your application, unless
you're doing something super unusual.
> 2 - Is there any way to get transparent compression with ext4? That
> would really, really rock and is, perhaps, one of the features that
> some users like me would greatly benefit from.
No, transparent compression is not a feature ext4 has. There has been
some thinking about creating a read-only compression feature,
primarily to speed up reading from slow devices, but full transparent
compression where users might want to seek to the middle of a
transparently compressioned file, and then writing into the middle of
said file, and making sure this is reliable even in the face of a
crash in the middle of doing such an update, etc., is something that
would be a large amount of work, for relatively little benefit, and so
I'd be surprised if that were ever to be supported in ext4.
Read-only compressed files is something that might happen at some
point, but while we have a design sketched out for it, no one is
currently actively working on that feature.
Cheers,
- Ted
prev parent reply other threads:[~2020-08-17 19:43 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-17 3:54 Generic questions about ext4 Rogério Brito
2020-08-17 19:43 ` Theodore Y. Ts'o [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200817194344.GA6755@mit.edu \
--to=tytso@mit.edu \
--cc=linux-ext4@vger.kernel.org \
--cc=rbrito@ime.usp.br \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).