. $LUSTRE/tests/test-framework.sh
init_test_env $@
. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+init_logging
+# bug 20670
+ALWAYS_EXCEPT="parallel_grouplock $PARALLEL_SCALE_EXCEPT"
+
+# common setup
#
+MACHINEFILE=${MACHINEFILE:-$TMP/$(basename $0 .sh).machines}
+clients=${CLIENTS:-$HOSTNAME}
+generate_machine_file $clients $MACHINEFILE || error "Failed to generate machine file"
+num_clients=$(get_node_count ${clients//,/ })
+
+
# compilbench
#
cbench_DIR=${cbench_DIR:-""}
METABENCH=${METABENCH:-$(which metabench 2> /dev/null || true)}
mbench_NFILES=${mbench_NFILES:-30400}
[ "$SLOW" = "no" ] && mbench_NFILES=10000
-MACHINEFILE=${MACHINEFILE:-$TMP/$(basename $0 .sh).machines}
# threads per client
mbench_THREADS=${mbench_THREADS:-4}
[ "$SLOW" = "no" ] && simul_REP=2
#
+# mib
+#
+MIB=${MIB:=$(which mib 2> /dev/null || true)}
+# threads per client
+mib_THREADS=${mib_THREADS:-2}
+mib_xferSize=${mib_xferSize:-1m}
+mib_xferLimit=${mib_xferLimit:-5000}
+mib_timeLimit=${mib_timeLimit:-300}
+
+#
+# MDTEST
+#
+MDTEST=${MDTEST:=$(which mdtest 2> /dev/null || true)}
+# threads per client
+mdtest_THREADS=${mdtest_THREADS:-2}
+mdtest_nFiles=${mdtest_nFiles:-"100000"}
+# We devide the files by number of core
+mdtest_nFiles=$((mdtest_nFiles/mdtest_THREADS/num_clients))
+mdtest_iteration=${mdtest_iteration:-1}
+
+#
# connectathon
#
cnt_DIR=${cnt_DIR:-""}
IOR=${IOR:-$(which IOR 2> /dev/null || true)}
# threads per client
ior_THREADS=${ior_THREADS:-2}
+ior_iteration=${ior_iteration:-1}
ior_blockSize=${ior_blockSize:-6} # Gb
+ior_xferSize=${ior_xferSize:-2m}
+ior_type=${ior_type:-POSIX}
ior_DURATION=${ior_DURATION:-30} # minutes
[ "$SLOW" = "no" ] && ior_DURATION=5
wdisjoint_REP=${wdisjoint_REP:-10000}
[ "$SLOW" = "no" ] && wdisjoint_REP=100
+#
+# parallel_grouplock
+#
+#
+PARALLEL_GROUPLOCK=${PARALLEL_GROUPLOCK:-$(which parallel_grouplock 2> /dev/null || true)}
+parallel_grouplock_MINTASKS=${parallel_grouplock_MINTASKS:-5}
+
build_test_filter
check_and_setup_lustre
MPI_RUNAS=${MPI_RUNAS:-"runas -u $MPI_USER_UID -g $MPI_USER_GID"}
$GSS_KRB5 && refresh_krb5_tgt $MPI_USER_UID $MPI_USER_GID $MPI_RUNAS
+
print_opts () {
local var
print_opts cbench_DIR cbench_IDIRS cbench_RUNS
[ x$cbench_DIR = x ] &&
- { skip "compilebench not found" && return; }
+ { skip_env "compilebench not found" && return; }
[ -e $cbench_DIR/compilebench ] || \
- { skip "No compilebench build" && return; }
+ { skip_env "No compilebench build" && return; }
local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }')
if [ $space -le $((680 * 1024 * cbench_IDIRS)) ]; then
cbench_IDIRS=$(( space / 680 / 1024))
[ $cbench_IDIRS = 0 ] && \
- skip "Need free space atleast 680 Mb, have $space" && return
+ skip_env "Need free space atleast 680 Mb, have $space" && return
log free space=$space, reducing initial dirs to $cbench_IDIRS
fi
mkdir -p $testdir
local savePWD=$PWD
- cd $cbench_DIR
+ cd $cbench_DIR
local cmd="./compilebench -D $testdir -i $cbench_IDIRS -r $cbench_RUNS --makej"
log "$cmd"
local rc=0
eval $cmd
rc=$?
-
+
cd $savePWD
[ $rc = 0 ] || error "compilebench failed: $rc"
rm -rf $testdir
test_metabench() {
[ x$METABENCH = x ] &&
- { skip "metabench not found" && return; }
-
- local clients=$CLIENTS
- [ -z $clients ] && clients=$(hostname)
-
- num_clients=$(get_node_count ${clients//,/ })
+ { skip_env "metabench not found" && return; }
# FIXME
# Need space estimation here.
- generate_machine_file $clients $MACHINEFILE || \
- error "can not generate machinefile $MACHINEFILE"
-
print_opts METABENCH clients mbench_NFILES mbench_THREADS
local testdir=$DIR/d0.metabench
# -k Cleanup. Remove the test directories.
local cmd="$METABENCH -w $testdir -c $mbench_NFILES -C -S -k"
echo "+ $cmd"
- mpi_run -np $((num_clients * $mbench_THREADS)) -machinefile ${MACHINEFILE} $cmd
+
+ # find out if we need to use srun by checking $SRUN_PARTITION
+ if [ "$SRUN_PARTITION" ]; then
+ $SRUN $SRUN_OPTIONS -D $testdir -w $clients -N $num_clients \
+ -n $((num_clients * mbench_THREADS)) -p $SRUN_PARTITION -- $cmd
+ else
+ mpi_run -np $((num_clients * $mbench_THREADS)) -machinefile ${MACHINEFILE} $cmd
+ fi
+
local rc=$?
if [ $rc != 0 ] ; then
error "metabench failed! $rc"
run_test metabench "metabench"
test_simul() {
- [ x$SIMUL = x ] &&
- { skip "simul not found" && return; }
-
- local clients=$CLIENTS
- [ -z $clients ] && clients=$(hostname)
+ if [ "$NFSCLIENT" ]; then
+ skip "skipped for NFSCLIENT mode"
+ return
+ fi
- local num_clients=$(get_node_count ${clients//,/ })
+ [ x$SIMUL = x ] &&
+ { skip_env "simul not found" && return; }
# FIXME
# Need space estimation here.
- generate_machine_file $clients $MACHINEFILE || \
- error "can not generate machinefile $MACHINEFILE"
-
print_opts SIMUL clients simul_REP simul_THREADS
local testdir=$DIR/d0.simul
local cmd="$SIMUL -d $testdir -n $simul_REP -N $simul_REP"
echo "+ $cmd"
- mpi_run -np $((num_clients * $simul_THREADS)) -machinefile ${MACHINEFILE} $cmd
+ # find out if we need to use srun by checking $SRUN_PARTITION
+ if [ "$SRUN_PARTITION" ]; then
+ $SRUN $SRUN_OPTIONS -D $testdir -w $clients -N $num_clients \
+ -n $((num_clients * simul_THREADS)) -p $SRUN_PARTITION -- $cmd
+ else
+ mpi_run -np $((num_clients * simul_THREADS)) -machinefile ${MACHINEFILE} $cmd
+ fi
local rc=$?
if [ $rc != 0 ] ; then
}
run_test simul "simul"
+test_mdtest() {
+ local type=${1:-"ssf"}
+
+ if [ "$NFSCLIENT" ]; then
+ skip "skipped for NFSCLIENT mode"
+ return
+ fi
+
+ [ x$MDTEST = x ] &&
+ { skip_env "mdtest not found" && return; }
+
+ # FIXME
+ # Need space estimation here.
+
+ print_opts MDTEST mdtest_iteration mdtest_THREADS mdtest_nFiles
+
+ local testdir=$DIR/d0.mdtest
+ mkdir -p $testdir
+ # mpi_run uses mpiuser
+ chmod 0777 $testdir
+
+ # -i # : repeat each test # times
+ # -d : test dir
+ # -n # : number of file/dir to create/stat/remove
+ # -u : each process create/stat/remove individually
+
+ local cmd="$MDTEST -d $testdir -i $mdtest_iteration -n $mdtest_nFiles"
+ [ $type = "fpp" ] && cmd="$cmd -u"
+
+ echo "+ $cmd"
+ # find out if we need to use srun by checking $SRUN_PARTITION
+ if [ "$SRUN_PARTITION" ]; then
+ $SRUN $SRUN_OPTIONS -D $testdir -w $clients -N $num_clients \
+ -n $((num_clients * mdtest_THREADS)) -p $SRUN_PARTITION -- $cmd
+ else
+ mpi_run -np $((num_clients * mdtest_THREADS)) -machinefile ${MACHINEFILE} $cmd
+ fi
+
+ local rc=$?
+ if [ $rc != 0 ] ; then
+ error "mdtest failed! $rc"
+ fi
+ rm -rf $testdir
+}
+
+test_mdtestssf() {
+ test_mdtest "ssf"
+}
+run_test mdtestssf "mdtestssf"
+
+test_mdtestfpp() {
+ test_mdtest "fpp"
+}
+run_test mdtestfpp "mdtestfpp"
+
test_connectathon() {
print_opts cnt_DIR cnt_NRUN
[ x$cnt_DIR = x ] &&
- { skip "connectathon dir not found" && return; }
+ { skip_env "connectathon dir not found" && return; }
[ -e $cnt_DIR/runtests ] || \
- { skip "No connectathon runtests found" && return; }
+ { skip_env "No connectathon runtests found" && return; }
local testdir=$DIR/d0.connectathon
mkdir -p $testdir
local savePWD=$PWD
cd $cnt_DIR
- # -f a quick functionality test
- # -a run basic, general, special, and lock tests
+ #
+ # cthon options (must be in this order)
+ #
# -N numpasses - will be passed to the runtests script. This argument
# is optional. It specifies the number of times to run
# through the tests.
-
- local cmd="./runtests -N $cnt_NRUN -a -f $testdir"
-
- log "$cmd"
-
- local rc=0
- eval $cmd
- rc=$?
+ #
+ # One of these test types
+ # -b basic
+ # -g general
+ # -s special
+ # -l lock
+ # -a all of the above
+ #
+ # -f a quick functionality test
+ #
+
+ tests="-b -g -s"
+ # Include lock tests unless we're running on nfsv4
+ local fstype=$(df -TP $testdir | awk 'NR==2 {print $2}')
+ echo "$testdir: $fstype"
+ if [[ $fstype != "nfs4" ]]; then
+ tests="$tests -l"
+ fi
+ echo "tests: $tests"
+ for test in $tests; do
+ local cmd="./runtests -N $cnt_NRUN $test -f $testdir"
+ local rc=0
+
+ log "$cmd"
+ eval $cmd
+ rc=$?
+ [ $rc = 0 ] || error "connectathon failed: $rc"
+ done
cd $savePWD
- [ $rc = 0 ] || error "connectathon failed: $rc"
rm -rf $testdir
}
run_test connectathon "connectathon"
test_ior() {
- [ x$IOR = x ] &&
- { skip "IOR not found" && return; }
-
- local clients=$CLIENTS
- [ -z $clients ] && clients=$(hostname)
+ local type=${1:="ssf"}
- local num_clients=$(get_node_count ${clients//,/ })
+ [ x$IOR = x ] &&
+ { skip_env "IOR not found" && return; }
local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }')
echo "+ $ior_blockSize * 1024 * 1024 * $num_clients * $ior_THREADS "
echo "+ $space * 9/10 / 1024 / 1024 / $num_clients / $ior_THREADS"
ior_blockSize=$(( space /2 /1024 /1024 / num_clients / ior_THREADS ))
[ $ior_blockSize = 0 ] && \
- skip "Need free space more than ($num_clients * $ior_THREADS )Gb: $((num_clients*ior_THREADS *1024 *1024*2)), have $space" && return
+ skip_env "Need free space more than ($num_clients * $ior_THREADS )Gb: $((num_clients*ior_THREADS *1024 *1024*2)), have $space" && return
echo "free space=$space, Need: $num_clients x $ior_THREADS x $ior_blockSize Gb (blockSize reduced to $ior_blockSize Gb)"
fi
-
- generate_machine_file $clients $MACHINEFILE || \
- error "can not generate machinefile $MACHINEFILE"
print_opts IOR ior_THREADS ior_DURATION MACHINEFILE
mkdir -p $testdir
# mpi_run uses mpiuser
chmod 0777 $testdir
-
- #
+ if [ "$NFSCLIENT" ]; then
+ setstripe_nfsserver $testdir -c -1 ||
+ { error "setstripe on nfsserver failed" && return 1; }
+ else
+ $LFS setstripe $testdir -c -1 ||
+ { error "setstripe failed" && return 2; }
+ fi
+ #
# -b N blockSize -- contiguous bytes to write per task (e.g.: 8, 4k, 2m, 1g)"
# -o S testFileName
# -t N transferSize -- size of transfer in bytes (e.g.: 8, 4k, 2m, 1g)"
# -r readFile -- read existing file"
# -T maxTimeDuration -- max time in minutes to run tests"
# -k keepFile -- keep testFile(s) on program exit
- local cmd="$IOR -a POSIX -b ${ior_blockSize}g -o $testdir/iorData -t 2m -v -w -r -T $ior_DURATION -k"
+
+ local cmd="$IOR -a $ior_type -b ${ior_blockSize}g -o $testdir/iorData -t $ior_xferSize -v -w -r -i $ior_iteration -T $ior_DURATION -k"
+ [ $type = "fpp" ] && cmd="$cmd -F"
echo "+ $cmd"
- mpi_run -np $((num_clients * $ior_THREADS)) -machinefile ${MACHINEFILE} $cmd
+ # find out if we need to use srun by checking $SRUN_PARTITION
+ if [ "$SRUN_PARTITION" ]; then
+ $SRUN $SRUN_OPTIONS -D $testdir -w $clients -N $num_clients \
+ -n $((num_clients * ior_THREADS)) -p $SRUN_PARTITION -- $cmd
+ else
+ mpi_run -np $((num_clients * $ior_THREADS)) -machinefile ${MACHINEFILE} $cmd
+ fi
local rc=$?
if [ $rc != 0 ] ; then
fi
rm -rf $testdir
}
-run_test ior "ior"
-
-test_cascading_rw() {
- [ x$CASC_RW = x ] &&
- { skip "cascading_rw not found" && return; }
- local clients=$CLIENTS
- [ -z $clients ] && clients=$(hostname)
+test_iorssf() {
+ test_ior "ssf"
+}
+run_test iorssf "iorssf"
+
+test_iorfpp() {
+ test_ior "fpp"
+}
+run_test iorfpp "iorfpp"
+
+test_mib() {
+ if [ "$NFSCLIENT" ]; then
+ skip "skipped for NFSCLIENT mode"
+ return
+ fi
+
+ [ x$MIB = x ] &&
+ { skip_env "MIB not found" && return; }
+
+ print_opts MIB mib_THREADS mib_xferSize mib_xferLimit mib_timeLimit MACHINEFILE
+
+ local testdir=$DIR/d0.mib
+ mkdir -p $testdir
+ # mpi_run uses mpiuser
+ chmod 0777 $testdir
+ $LFS setstripe $testdir -c -1 ||
+ { error "setstripe failed" && return 2; }
+ #
+ # -I Show intermediate values in output
+ # -H Show headers in output
+ # -L Do not issue new system calls after this many seconds
+ # -s Use system calls of this size
+ # -t test dir
+ # -l Issue no more than this many system calls
+ local cmd="$MIB -t $testdir -s $mib_xferSize -l $mib_xferLimit -L $mib_timeLimit -HI -p mib.$(date +%Y%m%d%H%M%S)"
+
+ echo "+ $cmd"
+ # find out if we need to use srun by checking $SRUN_PARTITION
+ if [ "$SRUN_PARTITION" ]; then
+ $SRUN $SRUN_OPTIONS -D $testdir -w $clients -N $num_clients \
+ -n $((num_clients * mib_THREADS)) -p $SRUN_PARTITION -- $cmd
+ else
+ mpi_run -np $((num_clients * mib_THREADS)) -machinefile ${MACHINEFILE} $cmd
+ fi
+
+ local rc=$?
+ if [ $rc != 0 ] ; then
+ error "mib failed! $rc"
+ fi
+ rm -rf $testdir
+}
+run_test mib "mib"
- num_clients=$(get_node_count ${clients//,/ })
+test_cascading_rw() {
+ if [ "$NFSCLIENT" ]; then
+ skip "skipped for NFSCLIENT mode"
+ return
+ fi
+
+ [ x$CASC_RW = x ] &&
+ { skip_env "cascading_rw not found" && return; }
# FIXME
# Need space estimation here.
- generate_machine_file $clients $MACHINEFILE || \
- error "can not generate machinefile $MACHINEFILE"
-
print_opts CASC_RW clients casc_THREADS casc_REP MACHINEFILE
local testdir=$DIR/d0.cascading_rw
# mpi_run uses mpiuser
chmod 0777 $testdir
- # -g: debug mode
+ # -g: debug mode
# -n: repeat test # times
local cmd="$CASC_RW -g -d $testdir -n $casc_REP"
run_test cascading_rw "cascading_rw"
test_write_append_truncate() {
- # location is lustre/tests dir
- if ! which write_append_truncate > /dev/null 2>&1 ; then
- skip "write_append_truncate not found"
+ if [ "$NFSCLIENT" ]; then
+ skip "skipped for NFSCLIENT mode"
return
fi
- local clients=$CLIENTS
- [ -z $clients ] && clients=$(hostname)
-
- local num_clients=$(get_node_count ${clients//,/ })
+ # location is lustre/tests dir
+ if ! which write_append_truncate > /dev/null 2>&1 ; then
+ skip_env "write_append_truncate not found"
+ return
+ fi
# FIXME
# Need space estimation here.
- generate_machine_file $clients $MACHINEFILE || \
- error "can not generate machinefile $MACHINEFILE"
-
local testdir=$DIR/d0.write_append_truncate
local file=$testdir/f0.wat
run_test write_append_truncate "write_append_truncate"
test_write_disjoint() {
- [ x$WRITE_DISJOINT = x ] &&
- { skip "write_disjoint not found" && return; }
-
- local clients=$CLIENTS
- [ -z $clients ] && clients=$(hostname)
+ if [ "$NFSCLIENT" ]; then
+ skip "skipped for NFSCLIENT mode"
+ return
+ fi
- local num_clients=$(get_node_count ${clients//,/ })
+ [ x$WRITE_DISJOINT = x ] &&
+ { skip_env "write_disjoint not found" && return; }
# FIXME
# Need space estimation here.
- generate_machine_file $clients $MACHINEFILE || \
- error "can not generate machinefile $MACHINEFILE"
-
print_opts WRITE_DISJOINT clients wdisjoint_THREADS wdisjoint_REP MACHINEFILE
local testdir=$DIR/d0.write_disjoint
mkdir -p $testdir
}
run_test write_disjoint "write_disjoint"
-equals_msg `basename $0`: test complete, cleaning up
+test_parallel_grouplock() {
+ if [ "$NFSCLIENT" ]; then
+ skip "skipped for NFSCLIENT mode"
+ return
+ fi
+
+ [ x$PARALLEL_GROUPLOCK = x ] &&
+ { skip "PARALLEL_GROUPLOCK not found" && return; }
+
+ print_opts clients parallel_grouplock_MINTASKS MACHINEFILE
+
+ local testdir=$DIR/d0.parallel_grouplock
+ mkdir -p $testdir
+ # mpi_run uses mpiuser
+ chmod 0777 $testdir
+
+ do_nodes $clients "lctl set_param llite.*.max_rw_chunk=0" ||
+ error "set_param max_rw_chunk=0 failed "
+
+ local cmd
+ local status=0
+ local subtest
+ for i in $(seq 12); do
+ subtest="-t $i"
+ local cmd="$PARALLEL_GROUPLOCK -g -v -d $testdir $subtest"
+ echo "+ $cmd"
+
+ mpi_run -np $parallel_grouplock_MINTASKS -machinefile ${MACHINEFILE} $cmd
+ local rc=$?
+ if [ $rc != 0 ] ; then
+ error_noexit "parallel_grouplock subtests $subtest failed! $rc"
+ else
+ echo "parallel_grouplock subtests $subtest PASS"
+ fi
+ let status=$((status + rc))
+ # clear debug to collect one log per one test
+ do_nodes $(comma_list $(nodes_list)) lctl clear
+ done
+ [ $status -eq 0 ] || error "parallel_grouplock status: $status"
+ rm -rf $testdir
+}
+run_test parallel_grouplock "parallel_grouplock"
+
+statahead_NUMMNTPTS=${statahead_NUMMNTPTS:-5}
+statahead_NUMFILES=${statahead_NUMFILES:-500000}
+
+cleanup_statahead () {
+ trap 0
+
+ local clients=$1
+ local mntpt_root=$2
+ local num_mntpts=$3
+
+ for i in $(seq 0 $num_mntpts);do
+ zconf_umount_clients $clients ${mntpt_root}$i ||
+ error_exit "Failed to umount lustre on ${mntpt_root}$i"
+ done
+}
+
+test_statahead () {
+ if [[ -n $NFSCLIENT ]]; then
+ skip "Statahead testing is not supported on NFS clients."
+ return 0
+ fi
+
+ [ x$MDSRATE = x ] &&
+ { skip_env "mdsrate not found" && return; }
+
+ print_opts MDSRATE clients statahead_NUMMNTPTS statahead_NUMFILES
+
+ # create large dir
+
+ # do not use default "d[0-9]*" dir name
+ # to avoid of rm $statahead_NUMFILES (500k) files in t-f cleanup
+ local dir=dstatahead
+ local testdir=$DIR/$dir
+
+ # cleanup only if dir exists
+ # cleanup only $statahead_NUMFILES number of files
+ # ignore the other files created by someone else
+ [ -d $testdir ] &&
+ mdsrate_cleanup $((num_clients * 32)) $MACHINEFILE $statahead_NUMFILES $testdir 'f%%d' --ignore
+
+ mkdir -p $testdir
+ # mpi_run uses mpiuser
+ chmod 0777 $testdir
+
+ local num_files=$statahead_NUMFILES
+
+ local IFree=$(inodes_available)
+ if [ $IFree -lt $num_files ]; then
+ num_files=$IFree
+ fi
+
+ cancel_lru_locks mdc
+
+ local cmd="${MDSRATE} ${MDSRATE_DEBUG} --mknod --dir $testdir --nfiles $num_files --filefmt 'f%%d'"
+ echo "+ $cmd"
+
+ mpi_run -np $((num_clients * 32)) -machinefile ${MACHINEFILE} $cmd
+
+ local rc=$?
+ if [ $rc != 0 ] ; then
+ error "mdsrate failed to create $rc"
+ return $rc
+ fi
+
+ local num_mntpts=$statahead_NUMMNTPTS
+ local mntpt_root=$TMP/mntpt/lustre
+ local mntopts=${MNTOPTSTATAHEAD:-$MOUNTOPT}
+
+ echo "Mounting $num_mntpts lustre clients starts on $clients"
+ trap "cleanup_statahead $clients $mntpt_root $num_mntpts" EXIT ERR
+ for i in $(seq 0 $num_mntpts); do
+ zconf_mount_clients $clients ${mntpt_root}$i "$mntopts" ||
+ error_exit "Failed to mount lustre on ${mntpt_root}$i on $clients"
+ done
+
+ do_rpc_nodes $clients cancel_lru_locks mdc
+
+ do_rpc_nodes $clients do_ls $mntpt_root $num_mntpts $dir
+
+ mdsrate_cleanup $((num_clients * 32)) $MACHINEFILE $num_files $testdir 'f%%d' --ignore
+
+ # use rm instead of rmdir because of
+ # testdir could contain the files created by someone else,
+ # or by previous run where is num_files prev > num_files current
+ rm -rf $testdir
+ cleanup_statahead $clients $mntpt_root $num_mntpts
+}
+
+run_test statahead "statahead test, multiple clients"
+
+complete $(basename $0) $SECONDS
check_and_cleanup_lustre
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true
+exit_status
git://git.whamcloud.com / fs / lustre-release.git / blobdiff