5 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
6 . $LUSTRE/tests/test-framework.sh
8 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
12 ALWAYS_EXCEPT="parallel_grouplock statahead $PARALLEL_SCALE_EXCEPT"
17 cbench_DIR=${cbench_DIR:-""}
18 cbench_IDIRS=${cbench_IDIRS:-4}
19 cbench_RUNS=${cbench_RUNS:-4} # FIXME: wiki page requirements is 30, do we really need 30 ?
21 if [ "$SLOW" = "no" ]; then
29 METABENCH=${METABENCH:-$(which metabench 2> /dev/null || true)}
30 mbench_NFILES=${mbench_NFILES:-30400}
31 [ "$SLOW" = "no" ] && mbench_NFILES=10000
32 MACHINEFILE=${MACHINEFILE:-$TMP/$(basename $0 .sh).machines}
34 mbench_THREADS=${mbench_THREADS:-4}
39 SIMUL=${SIMUL:=$(which simul 2> /dev/null || true)}
41 simul_THREADS=${simul_THREADS:-2}
42 simul_REP=${simul_REP:-20}
43 [ "$SLOW" = "no" ] && simul_REP=2
48 cnt_DIR=${cnt_DIR:-""}
49 cnt_NRUN=${cnt_NRUN:-10}
50 [ "$SLOW" = "no" ] && cnt_NRUN=2
55 CASC_RW=${CASC_RW:-$(which cascading_rw 2> /dev/null || true)}
57 casc_THREADS=${casc_THREADS:-2}
58 casc_REP=${casc_REP:-300}
59 [ "$SLOW" = "no" ] && casc_REP=10
64 IOR=${IOR:-$(which IOR 2> /dev/null || true)}
66 ior_THREADS=${ior_THREADS:-2}
67 ior_blockSize=${ior_blockSize:-6} # Gb
68 ior_DURATION=${ior_DURATION:-30} # minutes
69 [ "$SLOW" = "no" ] && ior_DURATION=5
72 # write_append_truncate
75 write_THREADS=${write_THREADS:-8}
76 write_REP=${write_REP:-10000}
77 [ "$SLOW" = "no" ] && write_REP=100
82 WRITE_DISJOINT=${WRITE_DISJOINT:-$(which write_disjoint 2> /dev/null || true)}
84 wdisjoint_THREADS=${wdisjoint_THREADS:-4}
85 wdisjoint_REP=${wdisjoint_REP:-10000}
86 [ "$SLOW" = "no" ] && wdisjoint_REP=100
92 PARALLEL_GROUPLOCK=${PARALLEL_GROUPLOCK:-$(which parallel_grouplock 2> /dev/null || true)}
93 parallel_grouplock_MINTASKS=${parallel_grouplock_MINTASKS:-5}
96 check_and_setup_lustre
98 get_mpiuser_id $MPI_USER
99 MPI_RUNAS=${MPI_RUNAS:-"runas -u $MPI_USER_UID -g $MPI_USER_GID"}
100 $GSS_KRB5 && refresh_krb5_tgt $MPI_USER_UID $MPI_USER_GID $MPI_RUNAS
109 echo "${var}=${!var}"
111 [ -e $MACHINEFILE ] && cat $MACHINEFILE
115 # 5 min * cbench_RUNS
119 # compile dir kernel-1 680MB
120 # required space 680MB * cbench_IDIRS = ~7 Gb
122 test_compilebench() {
123 print_opts cbench_DIR cbench_IDIRS cbench_RUNS
125 [ x$cbench_DIR = x ] &&
126 { skip_env "compilebench not found" && return; }
128 [ -e $cbench_DIR/compilebench ] || \
129 { skip_env "No compilebench build" && return; }
131 local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }')
132 if [ $space -le $((680 * 1024 * cbench_IDIRS)) ]; then
133 cbench_IDIRS=$(( space / 680 / 1024))
134 [ $cbench_IDIRS = 0 ] && \
135 skip_env "Need free space atleast 680 Mb, have $space" && return
137 log free space=$space, reducing initial dirs to $cbench_IDIRS
140 # t-f _base needs to be modifyed to set properly tdir
141 # for new "test_foo" functions names
142 # local testdir=$DIR/$tdir
143 local testdir=$DIR/d0.compilebench
148 local cmd="./compilebench -D $testdir -i $cbench_IDIRS -r $cbench_RUNS --makej"
157 [ $rc = 0 ] || error "compilebench failed: $rc"
160 run_test compilebench "compilebench"
163 [ x$METABENCH = x ] &&
164 { skip_env "metabench not found" && return; }
166 local clients=$CLIENTS
167 [ -z $clients ] && clients=$(hostname)
169 num_clients=$(get_node_count ${clients//,/ })
172 # Need space estimation here.
174 generate_machine_file $clients $MACHINEFILE || \
175 error "can not generate machinefile $MACHINEFILE"
177 print_opts METABENCH clients mbench_NFILES mbench_THREADS
179 local testdir=$DIR/d0.metabench
181 # mpi_run uses mpiuser
184 # -C Run the file creation tests.
185 # -S Run the file stat tests.
186 # -c nfile Number of files to be used in each test.
187 # -k Cleanup. Remove the test directories.
188 local cmd="$METABENCH -w $testdir -c $mbench_NFILES -C -S -k"
190 mpi_run -np $((num_clients * $mbench_THREADS)) -machinefile ${MACHINEFILE} $cmd
192 if [ $rc != 0 ] ; then
193 error "metabench failed! $rc"
197 run_test metabench "metabench"
201 { skip_env "simul not found" && return; }
203 local clients=$CLIENTS
204 [ -z $clients ] && clients=$(hostname)
206 local num_clients=$(get_node_count ${clients//,/ })
209 # Need space estimation here.
211 generate_machine_file $clients $MACHINEFILE || \
212 error "can not generate machinefile $MACHINEFILE"
214 print_opts SIMUL clients simul_REP simul_THREADS
216 local testdir=$DIR/d0.simul
218 # mpi_run uses mpiuser
221 # -n # : repeat each test # times
222 # -N # : repeat the entire set of tests # times
224 local cmd="$SIMUL -d $testdir -n $simul_REP -N $simul_REP"
227 mpi_run -np $((num_clients * $simul_THREADS)) -machinefile ${MACHINEFILE} $cmd
230 if [ $rc != 0 ] ; then
231 error "simul failed! $rc"
235 run_test simul "simul"
237 test_connectathon() {
238 print_opts cnt_DIR cnt_NRUN
241 { skip_env "connectathon dir not found" && return; }
243 [ -e $cnt_DIR/runtests ] || \
244 { skip_env "No connectathon runtests found" && return; }
246 local testdir=$DIR/d0.connectathon
252 # -f a quick functionality test
253 # -a run basic, general, special, and lock tests
254 # -N numpasses - will be passed to the runtests script. This argument
255 # is optional. It specifies the number of times to run
258 local cmd="./runtests -N $cnt_NRUN -a -f $testdir"
267 [ $rc = 0 ] || error "connectathon failed: $rc"
270 run_test connectathon "connectathon"
274 { skip_env "IOR not found" && return; }
276 local clients=$CLIENTS
277 [ -z $clients ] && clients=$(hostname)
279 local num_clients=$(get_node_count ${clients//,/ })
281 local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }')
282 echo "+ $ior_blockSize * 1024 * 1024 * $num_clients * $ior_THREADS "
283 if [ $((space / 2)) -le $(( ior_blockSize * 1024 * 1024 * num_clients * ior_THREADS)) ]; then
284 echo "+ $space * 9/10 / 1024 / 1024 / $num_clients / $ior_THREADS"
285 ior_blockSize=$(( space /2 /1024 /1024 / num_clients / ior_THREADS ))
286 [ $ior_blockSize = 0 ] && \
287 skip_env "Need free space more than ($num_clients * $ior_THREADS )Gb: $((num_clients*ior_THREADS *1024 *1024*2)), have $space" && return
289 echo "free space=$space, Need: $num_clients x $ior_THREADS x $ior_blockSize Gb (blockSize reduced to $ior_blockSize Gb)"
292 generate_machine_file $clients $MACHINEFILE || \
293 error "can not generate machinefile $MACHINEFILE"
295 print_opts IOR ior_THREADS ior_DURATION MACHINEFILE
297 local testdir=$DIR/d0.ior
299 # mpi_run uses mpiuser
301 $LFS setstripe $testdir -c -1
304 # -b N blockSize -- contiguous bytes to write per task (e.g.: 8, 4k, 2m, 1g)"
306 # -t N transferSize -- size of transfer in bytes (e.g.: 8, 4k, 2m, 1g)"
307 # -w writeFile -- write file"
308 # -r readFile -- read existing file"
309 # -T maxTimeDuration -- max time in minutes to run tests"
310 # -k keepFile -- keep testFile(s) on program exit
311 local cmd="$IOR -a POSIX -b ${ior_blockSize}g -o $testdir/iorData -t 2m -v -w -r -T $ior_DURATION -k"
314 mpi_run -np $((num_clients * $ior_THREADS)) -machinefile ${MACHINEFILE} $cmd
317 if [ $rc != 0 ] ; then
318 error "ior failed! $rc"
324 test_cascading_rw() {
325 if [ "$NFSCLIENT" ]; then
326 skip "skipped for NFSCLIENT mode"
331 { skip_env "cascading_rw not found" && return; }
333 local clients=$CLIENTS
334 [ -z $clients ] && clients=$(hostname)
336 num_clients=$(get_node_count ${clients//,/ })
339 # Need space estimation here.
341 generate_machine_file $clients $MACHINEFILE || \
342 error "can not generate machinefile $MACHINEFILE"
344 print_opts CASC_RW clients casc_THREADS casc_REP MACHINEFILE
346 local testdir=$DIR/d0.cascading_rw
348 # mpi_run uses mpiuser
352 # -n: repeat test # times
354 local cmd="$CASC_RW -g -d $testdir -n $casc_REP"
357 mpi_run -np $((num_clients * $casc_THREADS)) -machinefile ${MACHINEFILE} $cmd
360 if [ $rc != 0 ] ; then
361 error "cascading_rw failed! $rc"
365 run_test cascading_rw "cascading_rw"
367 test_write_append_truncate() {
368 # location is lustre/tests dir
369 if ! which write_append_truncate > /dev/null 2>&1 ; then
370 skip_env "write_append_truncate not found"
374 local clients=$CLIENTS
375 [ -z $clients ] && clients=$(hostname)
377 local num_clients=$(get_node_count ${clients//,/ })
380 # Need space estimation here.
382 generate_machine_file $clients $MACHINEFILE || \
383 error "can not generate machinefile $MACHINEFILE"
385 local testdir=$DIR/d0.write_append_truncate
386 local file=$testdir/f0.wat
388 print_opts clients write_REP write_THREADS MACHINEFILE
391 # mpi_run uses mpiuser
394 local cmd="write_append_truncate -n $write_REP $file"
397 mpi_run -np $((num_clients * $write_THREADS)) -machinefile ${MACHINEFILE} $cmd
400 if [ $rc != 0 ] ; then
401 error "write_append_truncate failed! $rc"
406 run_test write_append_truncate "write_append_truncate"
408 test_write_disjoint() {
409 [ x$WRITE_DISJOINT = x ] &&
410 { skip_env "write_disjoint not found" && return; }
412 local clients=$CLIENTS
413 [ -z $clients ] && clients=$(hostname)
415 local num_clients=$(get_node_count ${clients//,/ })
418 # Need space estimation here.
420 generate_machine_file $clients $MACHINEFILE || \
421 error "can not generate machinefile $MACHINEFILE"
423 print_opts WRITE_DISJOINT clients wdisjoint_THREADS wdisjoint_REP MACHINEFILE
424 local testdir=$DIR/d0.write_disjoint
426 # mpi_run uses mpiuser
429 local cmd="$WRITE_DISJOINT -f $testdir/file -n $wdisjoint_REP"
432 mpi_run -np $((num_clients * $wdisjoint_THREADS)) -machinefile ${MACHINEFILE} $cmd
435 if [ $rc != 0 ] ; then
436 error "write_disjoint failed! $rc"
440 run_test write_disjoint "write_disjoint"
442 test_parallel_grouplock() {
443 [ x$PARALLEL_GROUPLOCK = x ] &&
444 { skip "PARALLEL_GROUPLOCK not found" && return; }
446 local clients=$CLIENTS
447 [ -z $clients ] && clients=$(hostname)
449 local num_clients=$(get_node_count ${clients//,/ })
451 generate_machine_file $clients $MACHINEFILE || \
452 error "can not generate machinefile $MACHINEFILE"
454 print_opts clients parallel_grouplock_MINTASKS MACHINEFILE
456 local testdir=$DIR/d0.parallel_grouplock
458 # mpi_run uses mpiuser
461 do_nodes $clients "lctl set_param llite.*.max_rw_chunk=0" ||
462 error "set_param max_rw_chunk=0 failed "
467 for i in $(seq 12); do
469 local cmd="$PARALLEL_GROUPLOCK -g -v -d $testdir $subtest"
472 mpi_run -np $parallel_grouplock_MINTASKS -machinefile ${MACHINEFILE} $cmd
474 if [ $rc != 0 ] ; then
475 error_noexit "parallel_grouplock subtests $subtest failed! $rc"
477 echo "parallel_grouplock subtests $subtest PASS"
479 let status=$((status + rc))
480 # clear debug to collect one log per one test
481 do_nodes $(comma_list $(nodes_list)) lctl clear
483 [ $status -eq 0 ] || error "parallel_grouplock status: $status"
486 run_test parallel_grouplock "parallel_grouplock"
488 statahead_NUMMNTPTS=${statahead_NUMMNTPTS:-5}
489 statahead_NUMFILES=${statahead_NUMFILES:-500000}
491 cleanup_statahead () {
498 for i in $(seq 0 $num_mntpts);do
499 zconf_umount_clients $clients ${mntpt_root}$i ||
500 error_exit "Failed to umount lustre on ${mntpt_root}$i"
508 local dir=d0.statahead
509 # FIXME has to use DIR
510 local testdir=$DIR/$dir
514 local num_files=$statahead_NUMFILES
516 local IFree=$(inodes_available)
517 if [ $IFree -lt $num_files ]; then
523 log "createmany -o $testdir/f-%d $num_files"
524 createmany -o $testdir/$f-%d $num_files
527 if [ $rc != 0 ] ; then
528 error "createmany failed to create $rc"
532 local num_mntpts=$statahead_NUMMNTPTS
533 local mntpt_root=$TMP/mntpt/lustre
534 mntopts=${MNTOPTSTATAHEAD:-$MOUNTOPT}
536 local clients=$CLIENTS
537 [ -z $clients ] && clients=$(hostname)
539 echo "Mounting $num_mntpts lustre clients starts on $clients"
540 trap "cleanup_statahead $clients $mntpt_root $num_mntpts" EXIT ERR
541 for i in $(seq 0 $num_mntpts);do
542 zconf_mount_clients $clients ${mntpt_root}$i $mntopts ||
543 error_exit "Failed to mount lustre on ${mntpt_root}$i on $clients"
546 do_rpc_nodes $clients cancel_lru_locks mdc
548 do_rpc_nodes $clients do_ls $mntpt_root $num_mntpts $dir
550 cleanup_statahead $clients $mntpt_root $num_mntpts
553 run_test statahead "statahead test, multiple clients"
555 equals_msg `basename $0`: test complete, cleaning up
556 check_and_cleanup_lustre
557 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true