[PATCH] shared: new test to use up free inodes

[Eryu Guan]

On Thu, Mar 20, 2014 at 11:14:29AM +1100, Dave Chinner wrote:
> On Wed, Mar 19, 2014 at 05:27:49PM +0800, Eryu Guan wrote:
> > Stress test fs by using up all inodes and check fs.
> > 
> > Also a regression test for xfsprogs commit
> > d586858 xfs_repair: fix sibling pointer tests in verify_dir2_path()
> > 
> > Signed-off-by: Eryu Guan <eguan at redhat.com>
> > ---
> >  tests/shared/006     | 96 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> >  tests/shared/006.out |  2 ++
> >  tests/shared/group   |  1 +
> >  3 files changed, 99 insertions(+)
> >  create mode 100755 tests/shared/006
> >  create mode 100644 tests/shared/006.out
> > 
> > diff --git a/tests/shared/006 b/tests/shared/006
> > new file mode 100755
> > index 0000000..a3b13b6
> > --- /dev/null
> > +++ b/tests/shared/006
> > @@ -0,0 +1,96 @@
> > +#! /bin/bash
> > +# FS QA Test No. shared/006
> > +#
> > +# Stress test fs by using up all inodes and check fs.
> > +#
> > +# Also a regression test for xfsprogs commit
> > +# d586858 xfs_repair: fix sibling pointer tests in verify_dir2_path()
> > +#
> > +#-----------------------------------------------------------------------
> > +# Copyright (c) 2014 Red Hat Inc.  All Rights Reserved.
> > +#
> > +# This program is free software; you can redistribute it and/or
> > +# modify it under the terms of the GNU General Public License as
> > +# published by the Free Software Foundation.
> > +#
> > +# This program is distributed in the hope that it would be useful,
> > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > +# GNU General Public License for more details.
> > +#
> > +# You should have received a copy of the GNU General Public License
> > +# along with this program; if not, write the Free Software Foundation,
> > +# Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
> > +#-----------------------------------------------------------------------
> > +#
> > +
> > +seq=`basename $0`
> > +seqres=$RESULT_DIR/$seq
> > +echo "QA output created by $seq"
> > +
> > +here=`pwd`
> > +tmp=/tmp/$$
> > +status=1	# failure is the default!
> > +trap "_cleanup; exit \$status" 0 1 2 3 15
> > +
> > +_cleanup()
> > +{
> > +    cd /
> > +    rm -f $tmp.*
> > +}
> > +
> > +create_file()
> > +{
> > +	local dir=$1
> > +	local nr_file=$2
> > +	local prefix=$3
> > +	local i=0
> > +
> > +	while [ $i -lt $nr_file ]; do
> > +		touch $dir/${prefix}_${i}
> 
> echo -n > $dir/${prefix}_${i}
> 
> will create a zero length file without needing to fork/exec and so
> will have much lower overhead and create the files significantly
> faster.

This is better.

> 
> > +		let i=$i+1
> > +	done
> > +}
> > +
> > +# get standard environment, filters and checks
> > +. ./common/rc
> > +. ./common/filter
> > +
> > +# real QA test starts here
> > +_supported_fs ext4 ext3 ext2 xfs
> > +_supported_os Linux
> > +
> > +_require_scratch
> > +
> > +rm -f $seqres.full
> > +echo "Silence is golden"
> > +
> > +_scratch_mkfs_sized $((1024 * 1024 * 1024)) >>$seqres.full 2>&1
> > +_scratch_mount
> 
> If this is going to be a stress test, you should add a scale factor
> into this.

I don't have strong feeling to make it a stress test, I'd like to
remove the stress group to make the test a simple regression test.

> 
> > +
> > +i=0
> > +free_inode=`df -iP $SCRATCH_MNT | tail -1 | awk '{print $2}'`
> 
> $DF_PROG

OK.

> 
> > +loop=$((free_inode / 1000 + 1))
> 
> And probably a LOAD_FACTOR into this to scale parallelism.
> 
> > +mkdir -p $SCRATCH_MNT/testdir
> > +
> > +echo "Create $((loop * 1000)) files in $SCRATCH_MNT/testdir" >>$seqres.full
> > +while [ $i -lt $loop ]; do
> > +	create_file $SCRATCH_MNT/testdir 1000 $i >>$seqres.full 2>&1 &
> > +	let i=$i+1
> > +done
> > +wait
> 
> On XFS, that will create at least 500 threads creating 1000 inodes each
> all in the same directory. This doesn't give you any extra
> parallelism at all over just creating $free_inode files in a single
> directory with a single thread. Indeed, it will probably be slower
> due to the contention on the directory mutex.
> 
> If you want to scale this in terms of parallelism to keep the
> creation time down, each loop needs to write into a different
> directory. i.e. something like:
> 
> 
> echo "Create $((loop * 1000)) files in $SCRATCH_MNT/testdir" >>$seqres.full
> while [ $i -lt $loop ]; do
> 	mkdir -p $SCRATCH_MNT/testdir/$i
> 	create_file $SCRATCH_MNT/testdir/$i 1000 $i >>$seqres.full 2>&1 &
> 	let i=$i+1
> done
> wait
> 
> And even then I'd suggest that you'd be much better off with 10,000
> files to a sub-directory....

Will do.

> 
> > +# log inode status in $seqres.full for debug purpose
> > +echo "Inode status after taking all inodes" >>$seqres.full
> > +df -i $SCRATCH_MNT >>$seqres.full
> > +
> > +_check_scratch_fs
> > +
> > +# Check again after removing all the files
> > +rm -rf $SCRATCH_MNT/testdir
> 
> That can be parallelised as well when you have multiple subdirs:
> 
> for d in $SCRATCH_MNT/testdir/*; do
> 	rm -rf $d &
> done
> wait

Will do.

Thanks for the detailed review (as always)!

Eryu
> 
> Cheers,
> 
> Dave.
> -- 
> Dave Chinner
> david at fromorbit.com