From: Filipe Manana <fdmanana@gmail.com>
To: Anand Jain <anand.jain@oracle.com>
Cc: Josef Bacik <josef@toxicpanda.com>,
fstests <fstests@vger.kernel.org>,
linux-btrfs <linux-btrfs@vger.kernel.org>,
Eryu Guan <guaneryu@gmail.com>,
Johannes Thumshirn <johannes.thumshirn@wdc.com>
Subject: Re: [PATCH] fstests: delete btrfs/064 it makes no sense
Date: Wed, 30 Sep 2020 10:16:49 +0100 [thread overview]
Message-ID: <CAL3q7H6k4c9om0+of6yjib=OnNecnLvQMEi1n1NJWXK8L0MY5w@mail.gmail.com> (raw)
In-Reply-To: <a6c80503-df28-f25d-6437-657640bf8ade@oracle.com>
On Wed, Sep 30, 2020 at 5:14 AM Anand Jain <anand.jain@oracle.com> wrote:
>
> On 30/9/20 1:26 am, Josef Bacik wrote:
> > On 9/29/20 12:13 PM, Filipe Manana wrote:
> >> On Tue, Sep 29, 2020 at 5:02 PM Josef Bacik <josef@toxicpanda.com> wrote:
> >>>
> >>> On 9/29/20 11:55 AM, Filipe Manana wrote:
> >>>> On Tue, Sep 29, 2020 at 4:50 PM Anand Jain <anand.jain@oracle.com>
> >>>> wrote:
> >>>>>
> >>>>> btrfs/064 aimed to test balance and replace concurrency while the
> >>>>> stress
> >>>>> test is running in the background.
> >>>>>
> >>>>> However, as the balance and the replace operation are mutually
> >>>>> exclusive, so they can never run concurrently.
> >>>>
> >>>> And it's good to have a test that verifies that attempting to run them
> >>>> concurrently doesn't cause any problems, like crashes, memory leaks or
> >>>> some sort of filesystem corruption.
> >>>>
> >>>> For example btrfs/187, which I wrote sometime ago, tests that running
> >>>> send, balance and deduplication in parallel doesn't result in crashes,
> >>>> since in the past they were allowed to run concurrently.
> >>>>
> >>>> I see no point in removing the test, it's useful.
> >>>
> >>> My confusion was around whether this test was actually testing what we
> >>> think it should be testing. If this test was meant to make sure that
> >>> replace works while we've got load on the fs, then clearly it's not
> >>> doing what we think it's doing.
> >>
> >> Given that neither the test's description nor the changelog mention
> >> that it expects device replace and balance to be able to run
> >> concurrently,
> >> that errors are explicitly ignored and redirected to $seqres.full, and
> >> we don't do any sort of validation after device replace operations, it
> >> makes it clear to me it's a stress test.
> >>
> >
> > Sure but I spent a while looking at it when it was failing being very
> > confused. In my mind my snapshot-stress.sh is a stress test, because
> > its meant to run without errors. The changelog and description are
> > sufficiently vague enough that it appeared that Eryu meant to write a
> > test that actually did a replace and balance at the same time. The test
> > clearly isn't doing that, so we need to update the description so it's
> > clear that's what's going on. And then I wanted to make sure that we do
> > in fact have a test that stresses replace in these scenarios, because I
> > want to make sure we actually test replace as well.
> >
> > Not ripping it out is fine, but updating the description so I'm not
> > confused in a couple years when I trip over this again would be nice.
> > Thanks,
> >
>
> As of now, we have the following balance concurrency tests.
> -----
> 028 balance and unlink fsstress concurrency [1]
> 060 balance and subvol ops concurrency with fsstress [2]
> 061 balance and scrub concurrency with fsstress [2]
> 062 balance and defrag concurrency with fsstress [2]
> 063 balance and remount concurrency with fsstress [2]
> 064 balance and replace concurrency with fsstress [2]
> 177 balance and resize concurrency
No, 177 does not test balance and resize concurrency.
It tests balance when a swap file exists. And the resize happens
(starts and ends) before setting the swap file and before doing the
balance.
Thanks.
> 187 balance, send and dedupe concurrency
> 190 balance with qgroup
>
> [1]
> args=`_scale_fsstress_args -z \
> -f write=10 -f unlink=10 \
> -f creat=10 -f fsync=10 \
> -f fsync=10 -n 100000 -p 2 \
> -d $SCRATCH_MNT/stress_dir`
>
> [2]
> args=`_scale_fsstress_args -p 20 -n 100 $FSSTRESS_AVOID -d
> $SCRATCH_MNT/stressdir`
> -----
>
> 064 shall test balance with fsstress in the background. The replace
> thread is kept out with the early check of BTRFS_FS_EXCL_OP in the kernel.
> I am ok with the 064 headers updated, will send v2.
>
>
> Also, it turns out that this test case helped to find a btrfs-progs bug.
> Its patch [1] is sent to the ML.
> [1] btrfs-progs: fix return code for failed replace start
>
> Thanks, Anand
>
>
> > Josef
>
--
Filipe David Manana,
“Whether you think you can, or you think you can't — you're right.”
next prev parent reply other threads:[~2020-09-30 9:17 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-29 15:50 [PATCH] fstests: delete btrfs/064 it makes no sense Anand Jain
2020-09-29 15:55 ` Filipe Manana
2020-09-29 16:02 ` Josef Bacik
2020-09-29 16:13 ` Filipe Manana
2020-09-29 17:26 ` Josef Bacik
2020-09-30 4:14 ` Anand Jain
2020-09-30 9:16 ` Filipe Manana [this message]
2020-09-30 10:01 ` Anand Jain
2020-09-30 4:44 ` [PATCH v2] fstests: btrfs/064 add a comment to the test case header Anand Jain
2020-09-30 12:42 ` Josef Bacik
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='CAL3q7H6k4c9om0+of6yjib=OnNecnLvQMEi1n1NJWXK8L0MY5w@mail.gmail.com' \
--to=fdmanana@gmail.com \
--cc=anand.jain@oracle.com \
--cc=fstests@vger.kernel.org \
--cc=guaneryu@gmail.com \
--cc=johannes.thumshirn@wdc.com \
--cc=josef@toxicpanda.com \
--cc=linux-btrfs@vger.kernel.org \
/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).