X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fparallel-scale.sh;h=a7613e0d5843f005faa267b21389f5429288609c;hb=2410a014a1895cf90a00f6d54ec600bf7d42134a;hp=da4318c38b0d8e8364d5ccd515799086dd635436;hpb=be08548fa3e6bb505cb7ea6fd7a77ec50da1cbfb;p=fs%2Flustre-release.git diff --git a/lustre/tests/parallel-scale.sh b/lustre/tests/parallel-scale.sh index da4318c..a7613e0 100644 --- a/lustre/tests/parallel-scale.sh +++ b/lustre/tests/parallel-scale.sh @@ -7,12 +7,15 @@ LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)} init_test_env $@ . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh} +# bug 20670 +ALWAYS_EXCEPT="parallel_grouplock $PARALLEL_SCALE_EXCEPT" + # # compilbench # cbench_DIR=${cbench_DIR:-""} -cbench_IDIRS=${cbench_IDIRS:-10} -cbench_RUNS=${cbench_RUNS:-10} # FIXME: wiki page requirements is 30, do we really need 30 ? +cbench_IDIRS=${cbench_IDIRS:-4} +cbench_RUNS=${cbench_RUNS:-4} # FIXME: wiki page requirements is 30, do we really need 30 ? if [ "$SLOW" = "no" ]; then cbench_IDIRS=2 @@ -81,9 +84,20 @@ wdisjoint_THREADS=${wdisjoint_THREADS:-4} 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 +get_mpiuser_id $MPI_USER +MPI_RUNAS=${MPI_RUNAS:-"runas -u $MPI_USER_UID -g $MPI_USER_GID"} +$GSS_KRB5 && refresh_krb5_tgt $MPI_USER_UID $MPI_USER_GID $MPI_RUNAS + print_opts () { local var @@ -108,16 +122,16 @@ test_compilebench() { print_opts cbench_DIR cbench_IDIRS cbench_RUNS [ x$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 $cbench_IDIRS fi @@ -146,7 +160,7 @@ run_test compilebench "compilebench" test_metabench() { [ x$METABENCH = x ] && - { skip "metabench not found" && return; } + { skip_env "metabench not found" && return; } local clients=$CLIENTS [ -z $clients ] && clients=$(hostname) @@ -183,7 +197,7 @@ run_test metabench "metabench" test_simul() { [ x$SIMUL = x ] && - { skip "simul not found" && return; } + { skip_env "simul not found" && return; } local clients=$CLIENTS [ -z $clients ] && clients=$(hostname) @@ -223,10 +237,10 @@ 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 $testdir @@ -256,7 +270,7 @@ run_test connectathon "connectathon" test_ior() { [ x$IOR = x ] && - { skip "IOR not found" && return; } + { skip_env "IOR not found" && return; } local clients=$CLIENTS [ -z $clients ] && clients=$(hostname) @@ -269,7 +283,7 @@ test_ior() { 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 @@ -283,6 +297,7 @@ test_ior() { mkdir -p $testdir # mpi_run uses mpiuser chmod 0777 $testdir + $LFS setstripe $testdir -c -1 # # -b N blockSize -- contiguous bytes to write per task (e.g.: 8, 4k, 2m, 1g)" @@ -307,7 +322,7 @@ run_test ior "ior" test_cascading_rw() { [ x$CASC_RW = x ] && - { skip "cascading_rw not found" && return; } + { skip_env "cascading_rw not found" && return; } local clients=$CLIENTS [ -z $clients ] && clients=$(hostname) @@ -346,7 +361,7 @@ 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" + skip_env "write_append_truncate not found" return fi @@ -386,7 +401,7 @@ run_test write_append_truncate "write_append_truncate" test_write_disjoint() { [ x$WRITE_DISJOINT = x ] && - { skip "write_disjoint not found" && return; } + { skip_env "write_disjoint not found" && return; } local clients=$CLIENTS [ -z $clients ] && clients=$(hostname) @@ -418,6 +433,52 @@ test_write_disjoint() { } run_test write_disjoint "write_disjoint" +test_parallel_grouplock() { + [ x$PARALLEL_GROUPLOCK = x ] && + { skip "PARALLEL_GROUPLOCK not found" && return; } + + local clients=$CLIENTS + [ -z $clients ] && clients=$(hostname) + + local num_clients=$(get_node_count ${clients//,/ }) + + generate_machine_file $clients $MACHINEFILE || \ + error "can not generate machinefile $MACHINEFILE" + + print_opts clients parallel_grouplock_MINTASKS MACHINEFILE + + local testdir=$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" + equals_msg `basename $0`: test complete, cleaning up check_and_cleanup_lustre -[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true +[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true