From: Wido den Hollander <wido@widodh.nl>
To: Gregory Farnum <greg@inktank.com>
Cc: "Sébastien Han" <han.sebastien@gmail.com>,
ceph-devel <ceph-devel@vger.kernel.org>
Subject: Re: About sparse images
Date: Sun, 26 Aug 2012 21:53:34 +0200 [thread overview]
Message-ID: <503A7EBE.1060500@widodh.nl> (raw)
In-Reply-To: <649A15670E764B2E9397EDA1919205C7@inktank.com>
On 08/26/2012 09:42 PM, Gregory Farnum wrote:
> On Sunday, August 26, 2012 at 11:11 AM, Sébastien Han wrote:
>> Hi all!
>>
>> I know that ceph images (rbd) are sparse. It's good but I have some
>> interrogations about the ceph -s (-w) output. Let's say your create a
>> 10G image, map it, format it and mount it. Finally you filled up the
>> entire space available. You notice that the 'data' and 'used' space
>> had grown. Now delete the content of your image but **not** this
>> image, not now. You notice that nothing changed from the ceph output,
>> it's not surprising because of the sparse capability. At this point of
>> time the output from the command ceph is neither right nor wrong. It
>> only shows the space used by the sparse image, but it doesn't mean
>> that you have 'real' data in it, just a simple object allocation. It's
>> not so relevant, the only solution is to remove the image (obviously).
>> My feeling is that we should be careful about the ceph's output, it's
>> not a good indicator for 'real' data. Ok Ceph won't replace a
>> monitoring system but I'll be happy to trust the ceph output at 100%
>> :)
>
> The output of ceph -s is intended to describe the amount of space allocated by RADOS across the OSDs. As far as any part of the Ceph project knows, the objects in those RBD images are real data and can't be removed.
> If you're interested in removing them, subsets of the RBD clients do support TRIM operations, which will remove the objects themselves. Josh or somebody will have to speak about which clients, though.
> -Greg
>
As far as I know TRIM / Discard only works with KVM and SCSI disks,
VirtIO still seems to lack this functionality, not sure about kernel RBD.
On the filesystem side, ext4 with the "discard" mount option does this.
But it's important (like Greg already said) to note that RBD/RADOS can't
know which object is still important and which isn't. When you unlink a
file the filesystem doesn't zero that particular block, it just clears
all references to it so the freed up space can be used again.
Wido
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2012-08-26 19:53 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-08-26 18:11 About sparse images Sébastien Han
2012-08-26 18:38 ` Smart Weblications GmbH
2012-08-26 19:42 ` Gregory Farnum
2012-08-26 19:53 ` Wido den Hollander [this message]
2012-08-27 19:56 ` Josh Durgin
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=503A7EBE.1060500@widodh.nl \
--to=wido@widodh.nl \
--cc=ceph-devel@vger.kernel.org \
--cc=greg@inktank.com \
--cc=han.sebastien@gmail.com \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.