[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[linux-lvm] Draft Snapshot Regression Test


I just sent the below snapshot regression test to the XFS mailing list

It is designed to run from the XFS test harness, but that it should be easy to convert to be simply run by hand.

I have only tried it on a LVM 1.0.3 / xfs 1.1 / 2.4.19pre4-aa kernel and it fails within 10 iterations every time.

I have seen a couple of successes, but only when I set the iteration count very low, like 2 or 3.

On one occasion, the lvcreate locked up so tight, I had to reboot my server to get that FS unlocked.  (i.e. cycle power because the shutdown command would not run.)

There are a few environmental variables at the beginning you can set to control a couple of things.

The script only uses the $SCRATCH_DEV and $SCRATCH_MNT variables from the xfs test harness, so you could simply uncomment the currently commented out assignments.  The lines with _notrun in them can probably be deleted, unless you have a similar test harness.

I have not tested any of the TODO lines, but you may or may not need them in your test environments.

I copied the license stuff from another SGI test script, but I changed it to my company instead of SGI.

Feel free to make changes to this as required.

The script is below my signature.

BTW: This is mostly lvm specific, not xfs specific, so you can likely extend it to other filesystem types fairly easily.

Greg Freemyer
Internet Engineer
Deployment and Integration Specialist
Compaq ASE - Tru64 v4, v5
Compaq Master ASE - SAN Architect
The Norcross Group

#! /bin/sh
# XFS QA Test No. 068
# $Id: 1.1 $
# Test LVM snapshot creation
# The timing and placement of kills and waits is particularily sensitive.
#  Don't change them unless you want to spend some time getting it right again.
# Copyright (c) 2000, 2002 The Norcross Group, Inc.  All Rights Reserved.
# This program is free software; you can redistribute it and/or modify it
# under the terms of version 2 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
# Further, this software is distributed without any warranty that it is
# free of the rightful claim of any third person regarding infringement
# or the like.  Any license provided herein, whether implied or
# otherwise, applies only to this software file.  Patent licenses, if
# any, provided herein do not apply to combinations of this program with
# other software, or any other product whatsoever.
# You should have received a copy of the GNU General Public License along
# with this program; if not, write the Free Software Foundation, Inc., 59
# Temple Place - Suite 330, Boston MA 02111-1307, USA.
# Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
# Mountain View, CA  94043, or:
# http://www.sgi.com 
# For further information regarding this notice, see: 
# http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
# creator
owner=freemyer NorcrossGroup com

seq=`basename $0`
echo "QA output created by $seq"

#SCRATCH_DEV=/dev/xxxx          # Only needed if running by hand
#SCRATCH_MNT=/scratch           # Only needed if running by hand

status=1        # failure is the default!


       echo Cleanup beginning

       rm $tmp.running
       xfs_freeze -u $SCRATCH_MNT

       sleep 10      # Give the dd loop time to finish

       #Kill off any other possible stray stragglers that may be out there.
       fuser -k -m /scratch_snap/dummy     >/dev/null 2>&1
       fuser -k -m /$SCRATCH_MNT/dummy          >/dev/null 2>&1


       umount /scratch_snap
       rmdir /scratch_snap
       umount $SCRATCH_MNT

       lvremove -f $VG/scratch_snap
       lvremove -f $VG/scratch

#TODO   vgremove $VG

       rm -f $tmp.*        # if we ever use tmp files
       trap 0 1 2 3 15
       exit $status

trap "_cleanup" 0 1 2 3 15

# get standard environment, filters and checks
. ./common.rc
. ./common.filter

if [ -e /scratch_snap ]; then _notrun "This test requires that /scratch_snap not exist."; fi

mkdir /scratch_snap

#Verify we have the lvm user tools
LVM=`rpm -qa | grep 'lvm-'`

if [ x$LVM = "x" ]; then _notrun "LVM Must be installed to run this test"; fi

#Verify we have the a lvm enabled kernel
# TODO (This assumes lvm is a module.  What if it is linked?  I don't know how to check that.
lsmod | grep lvm-mod;
if [ $? != 0 ]; then _notrun "This test requires the LVM kernel module be present"; fi

# real QA test starts here

# Create a PV set from the scratch partition
# TODO # (I don't know if this is needed.and it is dangerous because it intentionally deletes the partition table!!! 
# TODO # dd if=/dev/zero of=$SCRATCH_DEV bs=512 count=1
# TODO # pvcreate $SCRATCH_DEV

# Create a VG from the PV
# TODO # vgcreate $VG $SCRATCH_DEV

# Create a LV in the VG to snapshot
# TODO # lvcreate -L 2G -n scratch $VG

# Mount the LV
mkdir /$SCRATCH_MNT > /dev/null 2&>1

mount $VG/scratch /$SCRATCH_MNT

# Create a large 64 Meg zero filled file on the LV
dd if=/dev/zero of=$SCRATCH_MNT/dummy bs=64k count=1000 

#setup an infinite loop to copy the large file, thus generating heavy i/o

touch $tmp.running

while [ -f $tmp.running ]
   dd if=$SCRATCH_MNT/dummy of=$SCRATCH_MNT/junk bs=64k > /dev/null 2>&1
   rm $SCRATCH_MNT/junk        # This forces metadata updates the next time around
done &


while [ $ii -le $ITERATIONS ]

       # echo $ii      Usefull if your are running interactive, but not from the xfs test scripts

       #if the VFS lock patch is present, the calls to xfs_freeze are redundant, but should cause no problems
       #       OPTIONAL
       xfs_freeze -f /$SCRATCH_MNT
       if [ $? != 0 ] ; then
               echo xfs_freeze -f /$SCRATCH_MNT failed
               lvcreate --snapshot --size 1G --name scratch_snap $VG/scratch > /dev/null 2>&1
               if [ $ret != 0 ] ; then
                       echo snapshot creation for /$SCRATCH_MNT failed with return code $ret
       ) &

       #if the Snapshot has not completed in ten minutes, kill it
                       # I have NOT figured out how to kill the sleep 600 before it exits naturally.
                       # This does not cause a problem, but it clutters the ps table.
               sleep 600
                       # The kill $TIMEOUT_shell_pid keeps the below from occuring
               echo Snapshot Lockup Occured on loop $ii
               xfs_freeze -u /$SCRATCH_MNT
               kill $$
       ) &

       wait $SNAPSHOT_shell_pid

       exec 2> /dev/null               # Send the shells stderr to /dev/null
       kill $TIMEOUT_shell_pid    #Cancel the timeout
       wait $TIMEOUT_shell_pid    # This causes consistent shell notification for some unknow reason
       exec 2>&1                       # Put it back to the same as stdout

       #if the VFS lock patch is present, the calls to xfs_freeze are redundant, but should cause no problems
       #       OPTIONAL
       xfs_freeze -u /$SCRATCH_MNT
       if [ $? != 0 ] ; then
               echo xfs_freeze -u /$SCRATCH_MNT failed
       #          MANDANTORY   (end)

       mount -t xfs -o ro,nouuid $VG/scratch_snap /scratch_snap
       if [ $? != 0 ] ; then
               echo mount for /scratch_snap failed
       umount /scratch_snap
       if [ $? != 0 ] ; then
              echo unmount for /scratch_snap failed
       lvremove -f $VG/scratch_snap > /dev/null 2>&1
       if [ $? != 0 ] ; then
              echo lvremove for /scratch_snap failed

ii=`expr $ii + 1`
       sleep $DELAY_BETWEEN_ITERATIONS # The VG seems to need time to stabalize between snapshots
                       # With LVM 1.0.3 and XFS 1.1, I have tried this at 3600 seconds and still had failures


# success, all done
exit 1   # _cleanup should exit, so we should never get here.

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]