xfs
[Top] [All Lists]

Re: [PATCH v2 1/3] xfstests/btrfs: add qgroup rescan stress test

To: Josef Bacik <jbacik@xxxxxx>
Subject: Re: [PATCH v2 1/3] xfstests/btrfs: add qgroup rescan stress test
From: Wang Shilong <wangsl.fnst@xxxxxxxxxxxxxx>
Date: Tue, 11 Mar 2014 09:19:39 +0800
Cc: xfs@xxxxxxxxxxx, linux-btrfs@xxxxxxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <531E15EA.9020903@xxxxxx>
References: <1394423051-17356-1-git-send-email-wangsl.fnst@xxxxxxxxxxxxxx> <531E15EA.9020903@xxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130612 Thunderbird/17.0.6
On 03/11/2014 03:43 AM, Josef Bacik wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 03/09/2014 11:44 PM, Wang Shilong wrote:
Test flow is to run fsstress after triggering quota rescan. the
ruler is simple, we just remove all files and directories, sync
filesystem and see if qgroup's ref and excl are nodesize.

Signed-off-by: Wang Shilong <wangsl.fnst@xxxxxxxxxxxxxx> ---
v1->v2: switch into new helper _run_btrfs_util_prog() ---
tests/btrfs/041     | 76
+++++++++++++++++++++++++++++++++++++++++++++++++++++
tests/btrfs/041.out |  3 +++ tests/btrfs/group   |  1 + 3 files
changed, 80 insertions(+) create mode 100644 tests/btrfs/041 create
mode 100644 tests/btrfs/041.out

So this doesn't pass for me but it's not because we're broken, it's
because the test is broken.  What you should do is run sync after you
do the rm, and then capture the output of btrfs qgroup show, and then
do the rescan and redo qgroup show and make sure the numbers match.  I
did this in my qgroup test, you can look there as an example.  Thanks,
Let's take a look at my test flows:

step1: write some random data to fs tree and then create a snapshot of fs tree.

step2: also try to write some random data to newly created snapshot.

step3: enable quota and wait quota rescan finished.(btrfs quota rescan -w)

step4: remove all files and sync btrfs filesystem.

So i did a simple test that if we don't create snapshot here, i get qgroup accounting right here, so i think it is realted to qgroup accounting that differs 'refer' and 'excl'.

Thanks,
Wang

Josef

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
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=/ZGH
-----END PGP SIGNATURE-----


<Prev in Thread] Current Thread [Next in Thread>