ALWAYS_EXCEPT="$SANITY_HSM_EXCEPT "
if $SHARED_KEY; then
# bug number for skipped tests: LU-9795 LU-9795
- ALWAYS_EXCEPT+=" 13 402b"
+ ALWAYS_EXCEPT+=" 13 402b "
# UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
fi
+# Skip tests for PPC that fail frequently
+if [[ $(uname -m) = ppc64 ]]; then
+ # bug number: LU-12251 LU-12251 LU-12251 LU-12251 LU-12251 LU-12251
+ ALWAYS_EXCEPT+=" 1a 1b 1d 1e 12c 12f "
+ # bug number: LU-12251 LU-12251 LU-12251 LU-12251 LU-12251 LU-12251
+ ALWAYS_EXCEPT+=" 12g 12h 12m 12n 12o 12p "
+ # bug number: LU-12251 LU-12251 LU-12251 LU-12251 LU-12251 LU-12251
+ ALWAYS_EXCEPT+=" 12q 21 22 23 24a 24b "
+ # bug number: LU-12251 LU-12251 LU-12251 LU-12251 LU-12251 LU-12251
+ ALWAYS_EXCEPT+=" 24d 24e 24f 25b 30c 37 "
+ # bug number: LU-12251 LU-12251 LU-12251 LU-12251 LU-12251 LU-12251
+ ALWAYS_EXCEPT+=" 57 58 90 110b 111b 113 "
+ # bug number: LU-12251 LU-12251 LU-12251 LU-12251 LU-12251 LU-12251
+ ALWAYS_EXCEPT+=" 222b 222d 228 260a 260b 260c "
+ # bug number: LU-12252 LU-12252 LU-12252 LU-12252 LU-12252 LU-12252
+ ALWAYS_EXCEPT+=" 220A 220a 221 222a 222c 223a "
+ # bug number: LU-12252 LU-12252 LU-12252 LU-12252 LU-12252 LU-12252
+ ALWAYS_EXCEPT+=" 223b 224A 224a 226 227 600"
+ # bug number: LU-12252 LU-12252 LU-12252 LU-12252 LU-12252 LU-12252
+ ALWAYS_EXCEPT+=" 601 602 603 604 605 "
+fi
+
build_test_filter
[ -n "$FILESET" ] && skip "Not functional for FILESET set"
OPENFILE=${OPENFILE:-openfile}
-MMAP_CAT=${MMAP_CAT:-mmap_cat}
MOUNT_2=${MOUNT_2:-"yes"}
FAIL_ON_ERROR=false
# Exception is the test which need two separate nodes
CLIENT2=${CLIENT2:-$CLIENT1}
-#
-# In order to test multiple remote HSM agents, a new facet type named "AGT" and
-# the following associated variables are added:
-#
-# AGTCOUNT: number of agents
-# AGTDEV{N}: target HSM mount point (root path of the backend)
-# agt{N}_HOST: hostname of the agent agt{N}
-# SINGLEAGT: facet of the single agent
-#
-# The number of agents is initialized as the number of remote client nodes.
-# By default, only single copytool is started on a remote client/agent. If there
-# was no remote client, then the copytool will be started on the local client.
-#
-init_agt_vars() {
- local n
- local agent
-
- export AGTCOUNT=${AGTCOUNT:-$((CLIENTCOUNT - 1))}
- [[ $AGTCOUNT -gt 0 ]] || AGTCOUNT=1
-
- export SHARED_DIRECTORY=${SHARED_DIRECTORY:-$TMP}
- if [[ $CLIENTCOUNT -gt 1 ]] &&
- ! check_shared_dir $SHARED_DIRECTORY $CLIENTS; then
- skip_env "SHARED_DIRECTORY should be accessible"\
- "on all client nodes"
- exit 0
- fi
-
- # We used to put the HSM archive in $SHARED_DIRECTORY but that
- # meant NFS issues could hose sanity-hsm sessions. So now we
- # use $TMP instead.
- for n in $(seq $AGTCOUNT); do
- eval export AGTDEV$n=\$\{AGTDEV$n:-"$TMP/arc$n"\}
- agent=CLIENT$((n + 1))
- if [[ -z "${!agent}" ]]; then
- [[ $CLIENTCOUNT -eq 1 ]] && agent=CLIENT1 ||
- agent=CLIENT2
- fi
- eval export agt${n}_HOST=\$\{agt${n}_HOST:-${!agent}\}
- done
-
- export SINGLEAGT=${SINGLEAGT:-agt1}
-
- export HSMTOOL=${HSMTOOL:-"lhsmtool_posix"}
- export HSMTOOL_VERBOSE=${HSMTOOL_VERBOSE:-""}
- export HSMTOOL_UPDATE_INTERVAL=${HSMTOOL_UPDATE_INTERVAL:=""}
- export HSMTOOL_EVENT_FIFO=${HSMTOOL_EVENT_FIFO:=""}
- export HSMTOOL_TESTDIR
-
- HSM_ARCHIVE_NUMBER=2
-
- # The test only support up to 10 MDTs
- MDT_PREFIX="mdt.$FSNAME-MDT000"
- HSM_PARAM="${MDT_PREFIX}0.hsm"
-
- # archive is purged at copytool setup
- HSM_ARCHIVE_PURGE=true
-
- # Don't allow copytool error upon start/setup
- HSMTOOL_NOERROR=false
-}
-
-# Get the backend root path for the given agent facet.
-copytool_device() {
- local facet=$1
- local dev=AGTDEV$(facet_number $facet)
-
- echo -n ${!dev}
-}
-
-get_mdt_devices() {
- local mdtno
- # get MDT device for each mdc
- for mdtno in $(seq 1 $MDSCOUNT); do
- local idx=$(($mdtno - 1))
- MDT[$idx]=$($LCTL get_param -n \
- mdc.$FSNAME-MDT000${idx}-mdc-*.mds_server_uuid |
- awk '{gsub(/_UUID/,""); print $1}' | head -n1)
- done
-}
-
search_copytools() {
local hosts=${1:-$(facet_active_host $SINGLEAGT)}
do_nodesv $hosts "libtool execute pgrep -x $HSMTOOL"
}
-kill_copytools() {
- local hosts=${1:-$(facet_active_host $SINGLEAGT)}
-
- echo "Killing existing copytools on $hosts"
- do_nodesv $hosts "libtool execute killall -q $HSMTOOL" || true
- copytool_continue "$hosts"
-}
-
wait_copytools() {
local hosts=${1:-$(facet_active_host $SINGLEAGT)}
local wait_timeout=200
local wait_start=$SECONDS
local wait_end=$((wait_start + wait_timeout))
- local sleep_time=100000 # 0.1 second
+ local sleep_time=1
while ((SECONDS < wait_end)); do
if ! search_copytools $hosts; then
fi
echo "copytools still running on $hosts"
- usleep $sleep_time
- [ $sleep_time -lt 32000000 ] && # 3.2 seconds
- sleep_time=$(bc <<< "$sleep_time * 2")
+ sleep $sleep_time
+ [ $sleep_time -lt 5 ] && sleep_time=$((sleep_time + 1))
done
# try to dump Copytool's stack
cmd="cat $test_dir/fifo > $test_dir/events &"
cmd+=" echo \\\$! > $test_dir/monitor_pid"
- if [[ $PDSH == *Rmrsh* ]]; then
- # This is required for pdsh -Rmrsh and its handling of remote
- # shells.
- # Regular ssh and pdsh -Rssh work fine without this
- # backgrounded subshell nonsense.
- (do_node $agent "$cmd") &
- export HSMTOOL_MONITOR_PDSH=$!
+ # This background subshell nonsense is required when pdsh/ssh decides
+ # to wait for the cat process to exit on the remote client
+ (do_node $agent "$cmd") &
+ export HSMTOOL_MONITOR_PDSH=$!
- # Slightly racy, but just making a best-effort to catch obvious
- # problems.
- sleep 1
- ps -p $HSMTOOL_MONITOR_PDSH > /dev/null ||
- error "Failed to start copytool monitor on $agent"
- else
- do_node $agent "$cmd"
- if [ $? != 0 ]; then
- error "Failed to start copytool monitor on $agent"
- fi
+ # Slightly racy, but just making a best-effort to catch obvious
+ # problems.
+ sleep 1
+ do_node $agent "stat $HSMTOOL_MONITOR_DIR/monitor_pid 2>&1 > /dev/null"
+ if [ $? != 0 ]; then
+ error "Failed to start copytool monitor on $agent"
fi
}
echo "Copytool is suspended on $agents"
}
-copytool_continue() {
- local agents=${1:-$(facet_active_host $SINGLEAGT)}
-
- do_nodesv $agents "libtool execute pkill -CONT -x $HSMTOOL" || return 0
- echo "Copytool is continued on $agents"
-}
-
copytool_remove_backend() {
local fid=$1
local be=$(do_facet $SINGLEAGT find "$(hsm_root)" -name $fid)
[[ -n $fid ]] && cmd+=" | grep '$fid'"
cmd+=" | egrep 'WAITING|STARTED'"
- wait_result $SINGLEMDS "$cmd" "" $timeout ||
+ wait_update_facet --verbose mds1 "$cmd" "" $timeout ||
error "requests did not complete"
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff