5 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
6 . $LUSTRE/tests/test-framework.sh
8 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
11 ALWAYS_EXCEPT="parallel_grouplock statahead $PARALLEL_SCALE_EXCEPT"
16 cbench_DIR=${cbench_DIR:-""}
17 cbench_IDIRS=${cbench_IDIRS:-4}
18 cbench_RUNS=${cbench_RUNS:-4} # FIXME: wiki page requirements is 30, do we really need 30 ?
20 if [ "$SLOW" = "no" ]; then
28 METABENCH=${METABENCH:-$(which metabench 2> /dev/null || true)}
29 mbench_NFILES=${mbench_NFILES:-30400}
30 [ "$SLOW" = "no" ] && mbench_NFILES=10000
31 MACHINEFILE=${MACHINEFILE:-$TMP/$(basename $0 .sh).machines}
33 mbench_THREADS=${mbench_THREADS:-4}
38 SIMUL=${SIMUL:=$(which simul 2> /dev/null || true)}
40 simul_THREADS=${simul_THREADS:-2}
41 simul_REP=${simul_REP:-20}
42 [ "$SLOW" = "no" ] && simul_REP=2
47 cnt_DIR=${cnt_DIR:-""}
48 cnt_NRUN=${cnt_NRUN:-10}
49 [ "$SLOW" = "no" ] && cnt_NRUN=2
54 CASC_RW=${CASC_RW:-$(which cascading_rw 2> /dev/null || true)}
56 casc_THREADS=${casc_THREADS:-2}
57 casc_REP=${casc_REP:-300}
58 [ "$SLOW" = "no" ] && casc_REP=10
63 IOR=${IOR:-$(which IOR 2> /dev/null || true)}
65 ior_THREADS=${ior_THREADS:-2}
66 ior_blockSize=${ior_blockSize:-6} # Gb
67 ior_DURATION=${ior_DURATION:-30} # minutes
68 [ "$SLOW" = "no" ] && ior_DURATION=5
71 # write_append_truncate
74 write_THREADS=${write_THREADS:-8}
75 write_REP=${write_REP:-10000}
76 [ "$SLOW" = "no" ] && write_REP=100
81 WRITE_DISJOINT=${WRITE_DISJOINT:-$(which write_disjoint 2> /dev/null || true)}
83 wdisjoint_THREADS=${wdisjoint_THREADS:-4}
84 wdisjoint_REP=${wdisjoint_REP:-10000}
85 [ "$SLOW" = "no" ] && wdisjoint_REP=100
91 PARALLEL_GROUPLOCK=${PARALLEL_GROUPLOCK:-$(which parallel_grouplock 2> /dev/null || true)}
92 parallel_grouplock_MINTASKS=${parallel_grouplock_MINTASKS:-5}
95 check_and_setup_lustre
97 get_mpiuser_id $MPI_USER
98 MPI_RUNAS=${MPI_RUNAS:-"runas -u $MPI_USER_UID -g $MPI_USER_GID"}
99 $GSS_KRB5 && refresh_krb5_tgt $MPI_USER_UID $MPI_USER_GID $MPI_RUNAS
108 echo "${var}=${!var}"
110 [ -e $MACHINEFILE ] && cat $MACHINEFILE
114 # 5 min * cbench_RUNS
118 # compile dir kernel-1 680MB
119 # required space 680MB * cbench_IDIRS = ~7 Gb
121 test_compilebench() {
122 print_opts cbench_DIR cbench_IDIRS cbench_RUNS
124 [ x$cbench_DIR = x ] &&
125 { skip_env "compilebench not found" && return; }
127 [ -e $cbench_DIR/compilebench ] || \
128 { skip_env "No compilebench build" && return; }
130 local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }')
131 if [ $space -le $((680 * 1024 * cbench_IDIRS)) ]; then
132 cbench_IDIRS=$(( space / 680 / 1024))
133 [ $cbench_IDIRS = 0 ] && \
134 skip_env "Need free space atleast 680 Mb, have $space" && return
136 log free space=$space, reducing initial dirs to $cbench_IDIRS
139 # t-f _base needs to be modifyed to set properly tdir
140 # for new "test_foo" functions names
141 # local testdir=$DIR/$tdir
142 local testdir=$DIR/d0.compilebench
147 local cmd="./compilebench -D $testdir -i $cbench_IDIRS -r $cbench_RUNS --makej"
156 [ $rc = 0 ] || error "compilebench failed: $rc"
159 run_test compilebench "compilebench"
162 [ x$METABENCH = x ] &&
163 { skip_env "metabench not found" && return; }
165 local clients=$CLIENTS
166 [ -z $clients ] && clients=$(hostname)
168 num_clients=$(get_node_count ${clients//,/ })
171 # Need space estimation here.
173 generate_machine_file $clients $MACHINEFILE || \
174 error "can not generate machinefile $MACHINEFILE"
176 print_opts METABENCH clients mbench_NFILES mbench_THREADS
178 local testdir=$DIR/d0.metabench
180 # mpi_run uses mpiuser
183 # -C Run the file creation tests.
184 # -S Run the file stat tests.
185 # -c nfile Number of files to be used in each test.
186 # -k Cleanup. Remove the test directories.
187 local cmd="$METABENCH -w $testdir -c $mbench_NFILES -C -S -k"
189 mpi_run -np $((num_clients * $mbench_THREADS)) -machinefile ${MACHINEFILE} $cmd
191 if [ $rc != 0 ] ; then
192 error "metabench failed! $rc"
196 run_test metabench "metabench"
200 { skip_env "simul not found" && return; }
202 local clients=$CLIENTS
203 [ -z $clients ] && clients=$(hostname)
205 local num_clients=$(get_node_count ${clients//,/ })
208 # Need space estimation here.
210 generate_machine_file $clients $MACHINEFILE || \
211 error "can not generate machinefile $MACHINEFILE"
213 print_opts SIMUL clients simul_REP simul_THREADS
215 local testdir=$DIR/d0.simul
217 # mpi_run uses mpiuser
220 # -n # : repeat each test # times
221 # -N # : repeat the entire set of tests # times
223 local cmd="$SIMUL -d $testdir -n $simul_REP -N $simul_REP"
226 mpi_run -np $((num_clients * $simul_THREADS)) -machinefile ${MACHINEFILE} $cmd
229 if [ $rc != 0 ] ; then
230 error "simul failed! $rc"
234 run_test simul "simul"
236 test_connectathon() {
237 print_opts cnt_DIR cnt_NRUN
240 { skip_env "connectathon dir not found" && return; }
242 [ -e $cnt_DIR/runtests ] || \
243 { skip_env "No connectathon runtests found" && return; }
245 local testdir=$DIR/d0.connectathon
251 # -f a quick functionality test
252 # -a run basic, general, special, and lock tests
253 # -N numpasses - will be passed to the runtests script. This argument
254 # is optional. It specifies the number of times to run
257 local cmd="./runtests -N $cnt_NRUN -a -f $testdir"
266 [ $rc = 0 ] || error "connectathon failed: $rc"
269 run_test connectathon "connectathon"
273 { skip_env "IOR not found" && return; }
275 local clients=$CLIENTS
276 [ -z $clients ] && clients=$(hostname)
278 local num_clients=$(get_node_count ${clients//,/ })
280 local space=$(df -P $DIR | tail -n 1 | awk '{ print $4 }')
281 echo "+ $ior_blockSize * 1024 * 1024 * $num_clients * $ior_THREADS "
282 if [ $((space / 2)) -le $(( ior_blockSize * 1024 * 1024 * num_clients * ior_THREADS)) ]; then
283 echo "+ $space * 9/10 / 1024 / 1024 / $num_clients / $ior_THREADS"
284 ior_blockSize=$(( space /2 /1024 /1024 / num_clients / ior_THREADS ))
285 [ $ior_blockSize = 0 ] && \
286 skip_env "Need free space more than ($num_clients * $ior_THREADS )Gb: $((num_clients*ior_THREADS *1024 *1024*2)), have $space" && return
288 echo "free space=$space, Need: $num_clients x $ior_THREADS x $ior_blockSize Gb (blockSize reduced to $ior_blockSize Gb)"
291 generate_machine_file $clients $MACHINEFILE || \
292 error "can not generate machinefile $MACHINEFILE"
294 print_opts IOR ior_THREADS ior_DURATION MACHINEFILE
296 local testdir=$DIR/d0.ior
298 # mpi_run uses mpiuser
300 $LFS setstripe $testdir -c -1
303 # -b N blockSize -- contiguous bytes to write per task (e.g.: 8, 4k, 2m, 1g)"
305 # -t N transferSize -- size of transfer in bytes (e.g.: 8, 4k, 2m, 1g)"
306 # -w writeFile -- write file"
307 # -r readFile -- read existing file"
308 # -T maxTimeDuration -- max time in minutes to run tests"
309 # -k keepFile -- keep testFile(s) on program exit
310 local cmd="$IOR -a POSIX -b ${ior_blockSize}g -o $testdir/iorData -t 2m -v -w -r -T $ior_DURATION -k"
313 mpi_run -np $((num_clients * $ior_THREADS)) -machinefile ${MACHINEFILE} $cmd
316 if [ $rc != 0 ] ; then
317 error "ior failed! $rc"
323 test_cascading_rw() {
324 if [ "$NFSCLIENT" ]; then
325 skip "skipped for NFSCLIENT mode"
330 { skip_env "cascading_rw not found" && return; }
332 local clients=$CLIENTS
333 [ -z $clients ] && clients=$(hostname)
335 num_clients=$(get_node_count ${clients//,/ })
338 # Need space estimation here.
340 generate_machine_file $clients $MACHINEFILE || \
341 error "can not generate machinefile $MACHINEFILE"
343 print_opts CASC_RW clients casc_THREADS casc_REP MACHINEFILE
345 local testdir=$DIR/d0.cascading_rw
347 # mpi_run uses mpiuser
351 # -n: repeat test # times
353 local cmd="$CASC_RW -g -d $testdir -n $casc_REP"
356 mpi_run -np $((num_clients * $casc_THREADS)) -machinefile ${MACHINEFILE} $cmd
359 if [ $rc != 0 ] ; then
360 error "cascading_rw failed! $rc"
364 run_test cascading_rw "cascading_rw"
366 test_write_append_truncate() {
367 # location is lustre/tests dir
368 if ! which write_append_truncate > /dev/null 2>&1 ; then
369 skip_env "write_append_truncate not found"
373 local clients=$CLIENTS
374 [ -z $clients ] && clients=$(hostname)
376 local num_clients=$(get_node_count ${clients//,/ })
379 # Need space estimation here.
381 generate_machine_file $clients $MACHINEFILE || \
382 error "can not generate machinefile $MACHINEFILE"
384 local testdir=$DIR/d0.write_append_truncate
385 local file=$testdir/f0.wat
387 print_opts clients write_REP write_THREADS MACHINEFILE
390 # mpi_run uses mpiuser
393 local cmd="write_append_truncate -n $write_REP $file"
396 mpi_run -np $((num_clients * $write_THREADS)) -machinefile ${MACHINEFILE} $cmd
399 if [ $rc != 0 ] ; then
400 error "write_append_truncate failed! $rc"
405 run_test write_append_truncate "write_append_truncate"
407 test_write_disjoint() {
408 [ x$WRITE_DISJOINT = x ] &&
409 { skip_env "write_disjoint not found" && return; }
411 local clients=$CLIENTS
412 [ -z $clients ] && clients=$(hostname)
414 local num_clients=$(get_node_count ${clients//,/ })
417 # Need space estimation here.
419 generate_machine_file $clients $MACHINEFILE || \
420 error "can not generate machinefile $MACHINEFILE"
422 print_opts WRITE_DISJOINT clients wdisjoint_THREADS wdisjoint_REP MACHINEFILE
423 local testdir=$DIR/d0.write_disjoint
425 # mpi_run uses mpiuser
428 local cmd="$WRITE_DISJOINT -f $testdir/file -n $wdisjoint_REP"
431 mpi_run -np $((num_clients * $wdisjoint_THREADS)) -machinefile ${MACHINEFILE} $cmd
434 if [ $rc != 0 ] ; then
435 error "write_disjoint failed! $rc"
439 run_test write_disjoint "write_disjoint"
441 test_parallel_grouplock() {
442 [ x$PARALLEL_GROUPLOCK = x ] &&
443 { skip "PARALLEL_GROUPLOCK not found" && return; }
445 local clients=$CLIENTS
446 [ -z $clients ] && clients=$(hostname)
448 local num_clients=$(get_node_count ${clients//,/ })
450 generate_machine_file $clients $MACHINEFILE || \
451 error "can not generate machinefile $MACHINEFILE"
453 print_opts clients parallel_grouplock_MINTASKS MACHINEFILE
455 local testdir=$DIR/d0.parallel_grouplock
457 # mpi_run uses mpiuser
460 do_nodes $clients "lctl set_param llite.*.max_rw_chunk=0" ||
461 error "set_param max_rw_chunk=0 failed "
466 for i in $(seq 12); do
468 local cmd="$PARALLEL_GROUPLOCK -g -v -d $testdir $subtest"
471 mpi_run -np $parallel_grouplock_MINTASKS -machinefile ${MACHINEFILE} $cmd
473 if [ $rc != 0 ] ; then
474 error_noexit "parallel_grouplock subtests $subtest failed! $rc"
476 echo "parallel_grouplock subtests $subtest PASS"
478 let status=$((status + rc))
479 # clear debug to collect one log per one test
480 do_nodes $(comma_list $(nodes_list)) lctl clear
482 [ $status -eq 0 ] || error "parallel_grouplock status: $status"
485 run_test parallel_grouplock "parallel_grouplock"
487 statahead_NUMMNTPTS=${statahead_NUMMNTPTS:-5}
488 statahead_NUMFILES=${statahead_NUMFILES:-500000}
490 cleanup_statahead () {
497 for i in $(seq 0 $num_mntpts);do
498 zconf_umount_clients $clients ${mntpt_root}$i ||
499 error_exit "Failed to umount lustre on ${mntpt_root}$i"
507 local dir=d0.statahead
508 # FIXME has to use DIR
509 local testdir=$DIR/$dir
513 local num_files=$statahead_NUMFILES
515 local IFree=$(inodes_available)
516 if [ $IFree -lt $num_files ]; then
522 log "createmany -o $testdir/f-%d $num_files"
523 createmany -o $testdir/$f-%d $num_files
526 if [ $rc != 0 ] ; then
527 error "createmany failed to create $rc"
531 local num_mntpts=$statahead_NUMMNTPTS
532 local mntpt_root=$TMP/mntpt/lustre
533 mntopts=${MNTOPTSTATAHEAD:-$MOUNTOPT}
535 local clients=$CLIENTS
536 [ -z $clients ] && clients=$(hostname)
538 echo "Mounting $num_mntpts lustre clients starts on $clients"
539 trap "cleanup_statahead $clients $mntpt_root $num_mntpts" EXIT ERR
540 for i in $(seq 0 $num_mntpts);do
541 zconf_mount_clients $clients ${mntpt_root}$i $mntopts ||
542 error_exit "Failed to mount lustre on ${mntpt_root}$i on $clients"
545 do_rpc_nodes $clients cancel_lru_locks mdc
547 do_rpc_nodes $clients do_ls $mntpt_root $num_mntpts $dir
549 cleanup_statahead $clients $mntpt_root $num_mntpts
552 run_test statahead "statahead test, multiple clients"
554 equals_msg `basename $0`: test complete, cleaning up
555 check_and_cleanup_lustre
556 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true