set -e
-# bug 5494 5493
-ALWAYS_EXCEPT="24 52 $RECOVERY_SMALL_EXCEPT"
+# bug 5493 LU2034
+ALWAYS_EXCEPT="52 60 $RECOVERY_SMALL_EXCEPT"
export MULTIOP=${MULTIOP:-multiop}
PTLDEBUG=${PTLDEBUG:--1}
# 1 2.5 2.5 4 4 (min)"
[ "$SLOW" = "no" ] && EXCEPT_SLOW="17 26a 26b 50 51 57"
+[ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
+# bug number for skipped test: LU-2194 LU-2547
+ ALWAYS_EXCEPT="$ALWAYS_EXCEPT 19b 24a 24b"
+
build_test_filter
# Allow us to override the setup if we already have a mounted system by
check_and_setup_lustre
assert_DIR
-rm -rf $DIR/[df][0-9]*
+rm -rf $DIR/d[0-9]* $DIR/f.${TESTSUITE}*
test_1() {
- drop_request "mcreate $DIR/f1" || return 1
- drop_reint_reply "mcreate $DIR/f2" || return 2
-}
-run_test 1 "mcreate: drop req, drop rep"
+ local f1="$DIR/$tfile"
+ local f2="$DIR/$tfile.2"
-test_2() {
- drop_request "tchmod 111 $DIR/f2" || return 1
- drop_reint_reply "tchmod 666 $DIR/f2" || return 2
-}
-run_test 2 "chmod: drop req, drop rep"
+ drop_request "mcreate $f1" ||
+ error_noexit "create '$f1': drop req"
-test_3() {
- drop_request "statone $DIR/f2" || return 1
- drop_reply "statone $DIR/f2" || return 2
-}
-run_test 3 "stat: drop req, drop rep"
+ drop_reint_reply "mcreate $f2" ||
+ error_noexit "create '$f2': drop rep"
+
+ drop_request "tchmod 111 $f2" ||
+ error_noexit "chmod '$f2': drop req"
+
+ drop_reint_reply "tchmod 666 $f2" ||
+ error_noexit "chmod '$f2': drop rep"
-SAMPLE_NAME=f0.recovery-small.junk
-SAMPLE_FILE=$TMP/$SAMPLE_NAME
-# make this big, else test 9 doesn't wait for bulk -- bz 5595
-dd if=/dev/urandom of=$SAMPLE_FILE bs=1M count=4
+ drop_request "statone $f2" ||
+ error_noexit "stat '$f2': drop req"
+
+ drop_reply "statone $f2" ||
+ error_noexit "stat '$f2': drop rep"
+}
+run_test 1 "create, chmod, stat: drop req, drop rep"
test_4() {
- do_facet client "cp $SAMPLE_FILE $DIR/$SAMPLE_NAME" || return 1
- drop_request "cat $DIR/$SAMPLE_NAME > /dev/null" || return 2
- drop_reply "cat $DIR/$SAMPLE_NAME > /dev/null" || return 3
+ local t=$DIR/$tfile
+ do_facet_create_file client $t 10K ||
+ error_noexit "Create file $t"
+
+ drop_request "cat $t > /dev/null" ||
+ error_noexit "Open request for $t file"
+
+ drop_reply "cat $t > /dev/null" ||
+ error_noexit "Open replay for $t file"
}
run_test 4 "open: drop req, drop rep"
-RENAMED_AGAIN=$DIR/f0.renamed-again
-
test_5() {
- drop_request "mv $DIR/$SAMPLE_NAME $DIR/$tfile-renamed" || return 1
- drop_reint_reply "mv $DIR/$tfile-renamed $RENAMED_AGAIN" || return 2
- do_facet client "checkstat -v $RENAMED_AGAIN" || return 3
+ local T=$DIR/$tfile
+ local R="$T-renamed"
+ local RR="$T-renamed-again"
+ do_facet_create_file client $T 10K ||
+ error_noexit "Create file $T"
+
+ drop_request "mv $T $R" ||
+ error_noexit "Rename $T"
+
+ drop_reint_reply "mv $R $RR" ||
+ error_noexit "Failed rename replay on $R"
+
+ do_facet client "checkstat -v $RR" ||
+ error_noexit "checkstat error on $RR"
+
+ do_facet client "rm $RR" ||
+ error_noexit "Can't remove file $RR"
}
run_test 5 "rename: drop req, drop rep"
-[ ! -e $RENAMED_AGAIN ] && cp $SAMPLE_FILE $RENAMED_AGAIN
-LINK1=$DIR/f0.link1
-LINK2=$DIR/f0.link2
-
test_6() {
- drop_request "mlink $RENAMED_AGAIN $LINK1" || return 1
- drop_reint_reply "mlink $RENAMED_AGAIN $LINK2" || return 2
-}
-run_test 6 "link: drop req, drop rep"
+ local T=$DIR/$tfile
+ local LINK1=$DIR/$tfile.link1
+ local LINK2=$DIR/$tfile.link2
+
+ do_facet_create_file client $T 10K ||
+ error_noexit "Create file $T"
+
+ drop_request "mlink $T $LINK1" ||
+ error_noexit "mlink request for $T"
+
+ drop_reint_reply "mlink $T $LINK2" ||
+ error_noexit "mlink reply for $T"
+
+ drop_request "munlink $LINK1" ||
+ error_noexit "munlink request for $T"
-[ ! -e $LINK1 ] && mlink $RENAMED_AGAIN $LINK1
-[ ! -e $LINK2 ] && mlink $RENAMED_AGAIN $LINK2
-test_7() {
- drop_request "munlink $LINK1" || return 1
- drop_reint_reply "munlink $LINK2" || return 2
+ drop_reint_reply "munlink $LINK2" ||
+ error_noexit "munlink reply for $T"
+
+ do_facet client "rm $T" ||
+ error_noexit "Can't remove file $T"
}
-run_test 7 "unlink: drop req, drop rep"
+run_test 6 "link, unlink: drop req, drop rep"
#bug 1423
test_8() {
#bug 1420
test_9() {
- remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+ remote_ost_nodsh && skip "remote OST with nodsh" && return 0
- pause_bulk "cp /etc/profile $DIR/$tfile" || return 1
- do_facet client "cp $SAMPLE_FILE $DIR/${tfile}.2" || return 2
- do_facet client "sync"
- do_facet client "rm $DIR/$tfile $DIR/${tfile}.2" || return 3
+ local t1=${tfile}.1
+ local t2=${tfile}.2
+ do_facet_random_file client $TMP/$tfile 1K ||
+ error_noexit "Create random file $TMP/$tfile"
+ # make this big, else test 9 doesn't wait for bulk -- bz 5595
+ do_facet_create_file client $TMP/$t1 4M ||
+ error_noexit "Create file $TMP/$t1"
+ do_facet client "cp $TMP/$t1 $DIR/$t1" ||
+ error_noexit "Can't copy to $DIR/$t1 file"
+ pause_bulk "cp $TMP/$tfile $DIR/$tfile" ||
+ error_noexit "Can't pause_bulk copy"
+ do_facet client "cp $TMP/$t1 $DIR/$t2" ||
+ error_noexit "Can't copy file"
+ do_facet client "sync"
+ do_facet client "rm $DIR/$tfile $DIR/$t2 $DIR/$t1" ||
+ error_noexit "Can't remove files"
+ do_facet client "rm $TMP/$t1 $TMP/$tfile"
}
run_test 9 "pause bulk on OST (bug 1420)"
#bug 1521
test_10() {
- do_facet client mcreate $DIR/$tfile || return 1
- drop_bl_callback "chmod 0777 $DIR/$tfile" || echo "evicted as expected"
- # wait for the mds to evict the client
- #echo "sleep $(($TIMEOUT*2))"
- #sleep $(($TIMEOUT*2))
- do_facet client touch $DIR/$tfile || echo "touch failed, evicted"
- do_facet client checkstat -v -p 0777 $DIR/$tfile || return 3
- do_facet client "munlink $DIR/$tfile"
+ do_facet client mcreate $DIR/$tfile ||
+ { error "mcreate failed: $?"; return 1; }
+ drop_bl_callback "chmod 0777 $DIR/$tfile" || echo "evicted as expected"
+ # wait for the mds to evict the client
+ #echo "sleep $(($TIMEOUT*2))"
+ #sleep $(($TIMEOUT*2))
+ do_facet client touch $DIR/$tfile || echo "touch failed, evicted"
+ do_facet client checkstat -v -p 0777 $DIR/$tfile ||
+ { error "client checkstat failed: $?"; return 3; }
+ do_facet client "munlink $DIR/$tfile"
+ # allow recovery to complete
+ client_up || client_up || sleep $TIMEOUT
}
run_test 10 "finish request on server after client eviction (bug 1521)"
#bug 2460
# wake up a thread waiting for completion after eviction
test_11(){
- do_facet client $MULTIOP $DIR/$tfile Ow || return 1
- do_facet client $MULTIOP $DIR/$tfile or || return 2
+ do_facet client $MULTIOP $DIR/$tfile Ow ||
+ { error "multiop write failed: $?"; return 1; }
+ do_facet client $MULTIOP $DIR/$tfile or ||
+ { error "multiop read failed: $?"; return 2; }
- cancel_lru_locks osc
+ cancel_lru_locks osc
- do_facet client $MULTIOP $DIR/$tfile or || return 3
- drop_bl_callback $MULTIOP $DIR/$tfile Ow || echo "evicted as expected"
+ do_facet client $MULTIOP $DIR/$tfile or ||
+ { error "multiop read failed: $?"; return 3; }
+ drop_bl_callback $MULTIOP $DIR/$tfile Ow || echo "evicted as expected"
- do_facet client munlink $DIR/$tfile || return 4
+ do_facet client munlink $DIR/$tfile ||
+ { error "munlink failed: $?"; return 4; }
+ # allow recovery to complete
+ client_up || client_up || sleep $TIMEOUT
}
run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
#b=2494
test_12(){
- $LCTL mark $MULTIOP $DIR/$tfile OS_c
- do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
- clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
- multiop_bg_pause $DIR/$tfile OS_c || return 1
- PID=$!
+ $LCTL mark $MULTIOP $DIR/$tfile OS_c
+ do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
+ clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
+ multiop_bg_pause $DIR/$tfile OS_c ||
+ { error "multiop failed: $?"; return 1; }
+ PID=$!
#define OBD_FAIL_MDS_CLOSE_NET 0x115
- kill -USR1 $PID
- echo "waiting for multiop $PID"
- wait $PID || return 2
- do_facet client munlink $DIR/$tfile || return 3
+ kill -USR1 $PID
+ echo "waiting for multiop $PID"
+ wait $PID || { error "wait for multiop faile: $?"; return 2; }
+ do_facet client munlink $DIR/$tfile ||
+ { error "client munlink failed: $?"; return 3; }
+ # allow recovery to complete
+ client_up || client_up || sleep $TIMEOUT
}
run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
# Bug 113, check that readdir lost recv timeout works.
test_13() {
- mkdir -p $DIR/$tdir || return 1
- touch $DIR/$tdir/newentry || return
+ mkdir -p $DIR/$tdir || { error "mkdir failed: $?"; return 1; }
+ touch $DIR/$tdir/newentry || { error "touch failed: $?"; return 2; }
# OBD_FAIL_MDS_READPAGE_NET|OBD_FAIL_ONCE
- do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
- ls $DIR/$tdir || return 3
- do_facet $SINGLEMDS "lctl set_param fail_loc=0"
- rm -rf $DIR/$tdir || return 4
+ do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
+ ls $DIR/$tdir || { error "ls failed: $?"; return 3; }
+ do_facet $SINGLEMDS "lctl set_param fail_loc=0"
+ rm -rf $DIR/$tdir || { error "remove test dir failed: $?"; return 4; }
}
run_test 13 "mdc_readpage restart test (bug 1138)"
}
test_16() {
- remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+ remote_ost_nodsh && skip "remote OST with nodsh" && return 0
- do_facet client cp $SAMPLE_FILE $DIR
- sync
- stop_read_ahead
+ do_facet_random_file client $TMP/$tfile 100K ||
+ { error_noexit "Create random file $TMP/$T" ; return 0; }
+ do_facet client "cp $TMP/$tfile $DIR/$tfile" ||
+ { error_noexit "Copy to $DIR/$tfile file" ; return 0; }
+ sync
+ stop_read_ahead
#define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | OBD_FAIL_ONCE
- do_facet ost1 "lctl set_param fail_loc=0x80000504"
- cancel_lru_locks osc
- # OST bulk will time out here, client resends
- do_facet client "cmp $SAMPLE_FILE $DIR/${SAMPLE_FILE##*/}" || return 1
- do_facet ost1 lctl set_param fail_loc=0
- # give recovery a chance to finish (shouldn't take long)
- sleep $TIMEOUT
- do_facet client "cmp $SAMPLE_FILE $DIR/${SAMPLE_FILE##*/}" || return 2
- start_read_ahead
+ do_facet ost1 "lctl set_param fail_loc=0x80000504"
+ cancel_lru_locks osc
+ # OST bulk will time out here, client resends
+ do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 1
+ do_facet ost1 lctl set_param fail_loc=0
+ # give recovery a chance to finish (shouldn't take long)
+ sleep $TIMEOUT
+ do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 2
+ start_read_ahead
}
run_test 16 "timeout bulk put, don't evict client (2732)"
remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+ local SAMPLE_FILE=$TMP/$tfile
+ do_facet_random_file client $SAMPLE_FILE 20K ||
+ { error_noexit "Create random file $SAMPLE_FILE" ; return 0; }
+
# With adaptive timeouts, bulk_get won't expire until adaptive_timeout_max
if at_is_enabled; then
at_max_saved=$(at_max_get ost1)
test_18a() {
[ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
+ do_facet_create_file client $TMP/$tfile 20K ||
+ { error_noexit "Create file $TMP/$tfile" ; return 0; }
+
do_facet client mkdir -p $DIR/$tdir
f=$DIR/$tdir/$tfile
error "$f: stripe_index $stripe_index != 1" && return
fi
- do_facet client cp $SAMPLE_FILE $f
+ do_facet client cp $TMP/$tfile $f
sync
local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
$LCTL --device $osc2dev deactivate || return 3
test_18b() {
remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+ do_facet_create_file client $TMP/$tfile 20K ||
+ { error_noexit "Create file $TMP/$tfile" ; return 0; }
+
do_facet client mkdir -p $DIR/$tdir
f=$DIR/$tdir/$tfile
error "$f: stripe_index $stripe_index != 0" && return
fi
- do_facet client cp $SAMPLE_FILE $f
+ do_facet client cp $TMP/$tfile $f
sync
ost_evict_client
# allow recovery to complete
test_18c() {
remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+ do_facet_create_file client $TMP/$tfile 20K ||
+ { error_noexit "Create file $TMP/$tfile" ; return 0; }
+
do_facet client mkdir -p $DIR/$tdir
f=$DIR/$tdir/$tfile
error "$f: stripe_index $stripe_index != 0" && return
fi
- do_facet client cp $SAMPLE_FILE $f
+ do_facet client cp $TMP/$tfile $f
sync
ost_evict_client
local BEFORE=`date +%s`
local EVICT
- mount_client $DIR2
+ mount_client $DIR2 || error "failed to mount $DIR2"
- do_facet client mcreate $DIR/$tfile || return 1
- drop_ldlm_cancel "chmod 0777 $DIR2"
+ # cancel cached locks from OST to avoid eviction from it
+ cancel_lru_locks osc
+
+ do_facet client "stat $DIR > /dev/null" ||
+ error "failed to stat $DIR: $?"
+ drop_ldlm_cancel "chmod 0777 $DIR2" ||
+ error "failed to chmod $DIR2"
umount_client $DIR2
- do_facet client "munlink $DIR/$tfile"
# let the client reconnect
- sleep 5
+ client_reconnect
EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state | \
awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
- [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] || error "no eviction"
+ [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
+ (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
+ error "no eviction: $EVICT before:$BEFORE")
}
run_test 19a "test expired_lock_main on mds (2867)"
local BEFORE=`date +%s`
local EVICT
- mount_client $DIR2
+ mount_client $DIR2 || error "failed to mount $DIR2: $?"
- do_facet client $MULTIOP $DIR/$tfile Ow || return 1
- drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow
- umount_client $DIR2
- do_facet client munlink $DIR/$tfile
+ # cancel cached locks from MDT to avoid eviction from it
+ cancel_lru_locks mdc
+
+ do_facet client $MULTIOP $DIR/$tfile Ow ||
+ error "failed to run multiop: $?"
+ drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow ||
+ error "failed to ldlm_cancel: $?"
+
+ umount_client $DIR2 || error "failed to unmount $DIR2: $?"
+ do_facet client munlink $DIR/$tfile ||
+ error "failed to unlink $DIR/$tfile: $?"
# let the client reconnect
- sleep 5
+ client_reconnect
EVICT=$(do_facet client $LCTL get_param osc.$FSNAME-OST*.state | \
awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
- [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] || error "no eviction"
+ [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
+ (do_facet client $LCTL get_param osc.$FSNAME-OST*.state;
+ error "no eviction: $EVICT before:$BEFORE")
}
run_test 19b "test expired_lock_main on ost (2867)"
+test_19c() {
+ local BEFORE=`date +%s`
+
+ mount_client $DIR2
+ $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0
+
+ mkdir -p $DIR1/$tfile
+ stat $DIR1/$tfile
+
+#define OBD_FAIL_PTLRPC_CANCEL_RESEND 0x516
+ do_facet mds $LCTL set_param fail_loc=0x80000516
+
+ touch $DIR2/$tfile/file1 &
+ PID1=$!
+ # let touch to get blocked on the server
+ sleep 2
+
+ wait $PID1
+ $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1
+ umount_client $DIR2
+
+ # let the client reconnect
+ sleep 5
+ EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
+ awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
+
+ [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
+}
+run_test 19c "check reconnect and lock resend do not trigger expired_lock_main"
+
test_20a() { # bug 2983 - ldlm_handle_enqueue cleanup
remote_ost_nodsh && skip "remote OST with nodsh" && return 0
}
run_test 23 "client hang when close a file after mds crash"
-test_24() { # bug 11710 details correct fsync() behavior
+test_24a() { # bug 11710 details correct fsync() behavior
remote_ost_nodsh && skip "remote OST with nodsh" && return 0
mkdir -p $DIR/$tdir
rc=$?
lctl set_param fail_loc=0x0
client_reconnect
- [ $rc -eq 0 ] && error_ignore 5494 "multiop didn't fail fsync: rc $rc" || true
+ [ $rc -eq 0 ] &&
+ error_ignore bz5494 "multiop didn't fail fsync: rc $rc" || true
}
-run_test 24 "fsync error (should return error)"
+run_test 24a "fsync error (should return error)"
wait_client_evicted () {
local facet=$1
local varsvc=${facet}_svc
wait_update $(facet_active_host $facet) \
- "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
- $((exports - 1)) $3
+ "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
+ $((exports - 1)) $3
}
+test_24b() {
+ remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+
+ dmesg -c > /dev/null
+ mkdir -p $DIR/$tdir
+ lfs setstripe $DIR/$tdir -s 0 -i 0 -c 1
+ cancel_lru_locks osc
+ multiop_bg_pause $DIR/$tdir/$tfile-1 Ow8192_yc ||
+ error "mulitop Ow8192_yc failed"
+
+ MULTI_PID1=$!
+ multiop_bg_pause $DIR/$tdir/$tfile-2 Ow8192_c ||
+ error "mulitop Ow8192_c failed"
+
+ MULTI_PID2=$!
+ ost_evict_client
+
+ kill -USR1 $MULTI_PID1
+ wait $MULTI_PID1
+ rc1=$?
+ kill -USR1 $MULTI_PID2
+ wait $MULTI_PID2
+ rc2=$?
+ lctl set_param fail_loc=0x0
+ client_reconnect
+ [ $rc1 -eq 0 -o $rc2 -eq 0 ] &&
+ error_ignore bz5494 "multiop didn't fail fsync: $rc1 or close: $rc2" ||
+ true
+
+ dmesg | grep "dirty page discard:" ||
+ error "no discarded dirty page found!"
+}
+run_test 24b "test dirty page discard due to client eviction"
+
test_26a() { # was test_26 bug 5921 - evict dead exports by pinger
# this test can only run from a client on a separate node.
remote_ost || { skip "local OST" && return 0; }
# fail abort so client will be new again
fail_abort $SINGLEMDS
client_up || error "reconnect failed"
+ wait_osc_import_state $SINGLEMDS ost FULL
return 0
}
run_test 29a "error adding new clients doesn't cause LBUG (bug 22273)"
rc=$?
echo writemany returned $rc
#these may fail because of eviction due to slow AST response.
- [ $rc -eq 0 ] || error_ignore 13652 "writemany returned rc $rc" || true
+ [ $rc -eq 0 ] ||
+ error_ignore bz13652 "writemany returned rc $rc" || true
}
run_test 50 "failover MDS under load"
SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
echo will failover at $SEQ
for i in $SEQ
- do
- echo failover in $i sec
- sleep $i
- facet_failover $SINGLEMDS
+ do
+ #echo failover in $i sec
+ log "test_$testnum: failover in $i sec"
+ sleep $i
+ facet_failover $SINGLEMDS
done
# client process should see no problems even though MDS went down
# and recovery was interrupted
sleep $TIMEOUT
kill -USR1 $CLIENT_PID
- wait $CLIENT_PID
+ wait $CLIENT_PID
rc=$?
echo writemany returned $rc
- [ $rc -eq 0 ] || error_ignore 13652 "writemany returned rc $rc" || true
+ [ $rc -eq 0 ] ||
+ error_ignore bz13652 "writemany returned rc $rc" || true
}
run_test 51 "failover MDS during recovery"
mkdir -p $DIR/$tdir
+ # Minimum pass speed is 2MBps
+ local ddtimeout=64
+ # LU-2887/LU-3089 - set min pass speed to 500KBps
+ [ "$(facet_fstype ost1)" = "zfs" ] && ddtimeout=256
+
# first dd should be finished quickly
$LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-1
- dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
+ dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
DDPID=$!
count=0
echo "step1: testing ......"
- while [ true ]; do
- if [ -z `ps x | awk '$1 == '$DDPID' { print $5 }'` ]; then break; fi
- count=$[count+1]
- if [ $count -gt 64 ]; then
- error "dd should be finished!"
- fi
- sleep 1
- done
+ while kill -0 $DDPID 2> /dev/null; do
+ let count++
+ if [ $count -gt $ddtimeout ]; then
+ error "dd should be finished!"
+ fi
+ sleep 1
+ done
echo "(dd_pid=$DDPID, time=$count)successful"
$LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-2
#define OBD_FAIL_OST_DROP_REQ 0x21d
do_facet ost1 lctl set_param fail_loc=0x0000021d
# second dd will be never finished
- dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
+ dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
DDPID=$!
count=0
echo "step2: testing ......"
- while [ $count -le 64 ]; do
- dd_name="`ps x | awk '$1 == '$DDPID' { print $5 }'`"
- if [ -z $dd_name ]; then
- ls -l $DIR/$tdir
- echo "debug: (dd_name=$dd_name, dd_pid=$DDPID, time=$count)"
- error "dd shouldn't be finished!"
- fi
- count=$[count+1]
- sleep 1
- done
+ while [ $count -le $ddtimeout ]; do
+ if ! kill -0 $DDPID 2> /dev/null; then
+ ls -l $DIR/$tdir
+ error "dd shouldn't be finished! (time=$count)"
+ fi
+ let count++
+ sleep 1
+ done
echo "(dd_pid=$DDPID, time=$count)successful"
#Recover fail_loc and dd will finish soon
do_facet ost1 lctl set_param fail_loc=0
count=0
echo "step3: testing ......"
- while [ true ]; do
- if [ -z `ps x | awk '$1 == '$DDPID' { print $5 }'` ]; then break; fi
- count=$[count+1]
- if [ $count -gt 500 ]; then
- error "dd should be finished!"
- fi
- sleep 1
- done
+ while kill -0 $DDPID 2> /dev/null; do
+ let count++
+ if [ $count -gt $((ddtimeout + 440)) ]; then
+ error "dd should be finished!"
+ fi
+ sleep 1
+ done
echo "(dd_pid=$DDPID, time=$count)successful"
- rm -rf $DIR/$tdir
+ rm -rf $DIR/$tdir
}
run_test 55 "ost_brw_read/write drops timed-out read/write request"
lctl set_param fail_loc=0
drop_bl_callback rm -f $DIR/$tfile
wait $pid
- do_facet client "df $DIR"
+ # the first 'df' could tigger the eviction caused by
+ # 'drop_bl_callback', and it's normal case.
+ # but the next 'df' should return successfully.
+ do_facet client "df $DIR" || do_facet client "df $DIR"
}
run_test 58 "Eviction in the middle of open RPC reply processing"
run_test 59 "Read cancel race on client eviction"
err17935 () {
- # we assume that all md changes are in the MDT0 changelog
- if [ $MDSCOUNT -gt 1 ]; then
- error_ignore 17935 $*
- else
- error $*
- fi
+ # we assume that all md changes are in the MDT0 changelog
+ if [ $MDSCOUNT -gt 1 ]; then
+ error_ignore bz17935 $*
+ else
+ error $*
+ fi
}
test_60() {
}
run_test 61 "Verify to not reuse orphan objects - bug 17025"
+# test_62 as seen it b2_1 please do not reuse test_62
+#test_62()
+#{
+# zconf_umount `hostname` $DIR
+# #define OBD_FAIL_PTLRPC_DELAY_IMP_FULL 0x516
+# lctl set_param fail_loc=0x516
+# mount_client $DIR
+#}
+#run_test 62 "Verify connection flags race - bug LU-1716"
+
check_cli_ir_state()
{
local NODE=${1:-$HOSTNAME}
local st
st=$(do_node $NODE "lctl get_param mgc.*.ir_state |
awk '/imperative_recovery:/ { print \\\$2}'")
- [ $st != ON -o $st != OFF ] ||
- error "Error state $st, must be ON or OFF"
+ [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
+ error "Error state $st, must be ENABLED or DISABLED"
echo -n $st
}
st=$(do_facet $target "lctl get_param -n $recovery_proc |
awk '/IR:/{ print \\\$2}'")
- [ $st != ON -o $st != OFF ] ||
- error "Error state $st, must be ON or OFF"
+ [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
+ error "Error state $st, must be ENABLED or DISABLED"
echo -n $st
}
test_100()
{
- do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
- { skip "MGS without IR support"; return 0; }
+ do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
+ { skip "MGS without IR support"; return 0; }
+
+ # MDT was just restarted in the previous test, make sure everything
+ # is all set.
+ local cnt=30
+ while [ $cnt -gt 0 ]; do
+ nidtbl_versions_match && break
+ sleep 1
+ cnt=$((cnt - 1))
+ done
+
+ # disable IR
+ set_ir_status disabled
- # disable IR
- set_ir_status disabled
+ local prev_ver=$(nidtbl_version_client client)
local saved_FAILURE_MODE=$FAILURE_MODE
[ $(facet_host mgs) = $(facet_host ost1) ] && FAILURE_MODE="SOFT"
fail ost1
# valid check
- nidtbl_versions_match &&
- error "version must differ due to IR disabled"
+ [ $(nidtbl_version_client client) -eq $prev_ver ] ||
+ error "version must not change due to IR disabled"
target_instance_match ost1 || error "instance mismatch"
# restore env
stop mds1
# We need this test because mds is like a client in IR context.
- start mds1 $MDSDEV1 || error "MDS should start w/o mgs"
+ start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS ||
+ error "MDS should start w/o mgs"
# start mgs and remount mds w/ ir
- start mgs $MGSDEV
+ start mgs $(mgsdevname) $MGS_MOUNT_OPTS
clients_up
# remount client so that fsdb will be created on the MGS
clients_up
local ir_state=$(check_target_ir_state ost1)
- [ $ir_state = "OFF" ] || error "ir status on ost1 should be OFF"
+ [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
+ error "ir status on ost1 should be DISABLED"
}
run_test 104 "IR: ost can disable IR voluntarily"
# make sure lustre mount at $rcli disabling IR
local ir_state=$(check_cli_ir_state $rcli)
- [ $ir_state = OFF ] || error "IR state must be OFF at $rcli"
+ [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
+ error "IR state must be DISABLED at $rcli"
# Since the client just mounted, its last_rcvd entry is not on disk.
# Send an RPC so exp_need_sync forces last_rcvd to commit this export
[ $(get_ir_status) = "partial" ] || error "MGS IR state must be partial"
fail ost1
- # make sure IR on ost1 is OFF
+ # make sure IR on ost1 is DISABLED
local ir_state=$(check_target_ir_state ost1)
- [ $ir_state = "OFF" ] || error "IR status on ost1 should be OFF"
+ [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
+ error "IR status on ost1 should be DISABLED"
# restore it
MOUNTOPT=$old_MOUNTOPT
[ $(get_ir_status) = "full" ] || error "MGS IR status must be full"
fail ost1
- # make sure IR on ost1 is ON
+ # make sure IR on ost1 is ENABLED
local ir_state=$(check_target_ir_state ost1)
- [ $ir_state = "ON" ] || error "IR status on ost1 should be OFF"
+ [ $ir_state = "ENABLED" -o $ir_state = "ON" ] ||
+ error "IR status on ost1 should be ENABLED"
return 0
}
run_test 105 "IR: NON IR clients support"
-complete $(basename $0) $SECONDS
+cleanup_106() {
+ trap 0
+ umount_client $DIR2
+ debugrestore
+}
+
+test_106() { # LU-1789
+ [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.50) ]] ||
+ { skip "Need MDS version at least 2.3.50"; return 0; }
+
+#define OBD_FAIL_MDC_LIGHTWEIGHT 0x805
+ $LCTL set_param fail_loc=0x805
+
+ debugsave
+ trap cleanup_106 EXIT
+
+ # enable lightweight flag on mdc connection
+ mount_client $DIR2
+
+ local MDS_NEXP=$(do_facet $SINGLEMDS \
+ lctl get_param -n mdt.${mds1_svc}.num_exports |
+ cut -d' ' -f2)
+ $LCTL set_param fail_loc=0
+
+ touch $DIR2/$tfile || error "failed to create empty file"
+ replay_barrier $SINGLEMDS
+
+ $LCTL set_param debug=console
+ $LCTL clear
+ facet_failover $SINGLEMDS
+
+ # lightweight connection must be evicted
+ touch -c $DIR2/$tfile || true
+ $LCTL dk $TMP/lustre-log-$TESTNAME.log
+ evicted=`awk '/This client was evicted by .*MDT0000/ {
+ print;
+ }' $TMP/lustre-log-$TESTNAME.log`
+ [ -z "$evicted" ] && error "lightweight client not evicted by mds"
+
+ # and all operations performed by lightweight client should be
+ # synchronous, so the file created before mds restart should be there
+ $CHECKSTAT -t file $DIR/$tfile || error "file not present"
+ rm -f $DIR/$tfile
+
+ cleanup_106
+}
+run_test 106 "lightweight connection support"
+
+test_107 () {
+ local CLIENT_PID
+ local close_pid
+
+ mkdir -p $DIR/$tdir
+ # OBD_FAIL_MDS_REINT_NET_REP 0x119
+ do_facet $SINGLEMDS lctl set_param fail_loc=0x119
+ multiop $DIR/$tdir D_c &
+ close_pid=$!
+ mkdir $DIR/$tdir/dir_106 &
+ CLIENT_PID=$!
+ do_facet $SINGLEMDS lctl set_param fail_loc=0
+ fail $SINGLEMDS
+
+ wait $CLIENT_PID || rc=$?
+ checkstat -t dir $DIR/$tdir/dir_106 || return 1
+
+ kill -USR1 $close_pid
+ wait $close_pid || return 2
+
+ return $rc
+}
+run_test 107 "drop reint reply, then restart MDT"
+
+test_110a () {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+ local remote_dir=$DIR/$tdir/remote_dir
+ local MDTIDX=1
+
+ mkdir -p $DIR/$tdir
+ drop_request "$LFS mkdir -i $MDTIDX $remote_dir" ||
+ error "lfs mkdir failed"
+ local diridx=$($GETSTRIPE -M $remote_dir)
+ [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
+
+ rm -rf $DIR/$tdir || error "rmdir failed"
+}
+run_test 110a "create remote directory: drop client req"
+
+test_110b () {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+ local remote_dir=$DIR/$tdir/remote_dir
+ local MDTIDX=1
+
+ mkdir -p $DIR/$tdir
+ drop_reint_reply "$LFS mkdir -i $MDTIDX $remote_dir" ||
+ error "lfs mkdir failed"
+
+ diridx=$($GETSTRIPE -M $remote_dir)
+ [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
+
+ rm -rf $DIR/$tdir || error "rmdir failed"
+}
+run_test 110b "create remote directory: drop Master rep"
+
+test_110c () {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+ local remote_dir=$DIR/$tdir/remote_dir
+ local MDTIDX=1
+
+ mkdir -p $DIR/$tdir
+ drop_update_reply $((MDTIDX + 1)) "$LFS mkdir -i $MDTIDX $remote_dir" ||
+ error "lfs mkdir failed"
+
+ diridx=$($GETSTRIPE -M $remote_dir)
+ [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
+
+ rm -rf $DIR/$tdir || error "rmdir failed"
+}
+run_test 110c "create remote directory: drop update rep on slave MDT"
+
+test_110d () {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+ local remote_dir=$DIR/$tdir/remote_dir
+ local MDTIDX=1
+
+ mkdir -p $DIR/$tdir
+ $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
+
+ drop_request "rm -rf $remote_dir" || error "rm remote dir failed"
+
+ rm -rf $DIR/$tdir || error "rmdir failed"
+
+ return 0
+}
+run_test 110d "remove remote directory: drop client req"
+
+test_110e () {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+ local remote_dir=$DIR/$tdir/remote_dir
+ local MDTIDX=1
+
+ mkdir -p $DIR/$tdir
+ $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
+ drop_reint_reply "rm -rf $remote_dir" || error "rm remote dir failed"
+
+ rm -rf $DIR/$tdir || error "rmdir failed"
+
+ return 0
+}
+run_test 110e "remove remote directory: drop master rep"
+
+test_110f () {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+ local remote_dir=$DIR/$tdir/remote_dir
+ local MDTIDX=1
+
+ mkdir -p $DIR/$tdir
+ $LFS mkdir -i $MDTIDX $remote_dir || error "lfs mkdir failed"
+ drop_update_reply $MDTIDX "rm -rf $remote_dir" ||
+ error "rm remote dir failed"
+
+ rm -rf $DIR/$tdir || error "rmdir failed"
+}
+run_test 110f "remove remote directory: drop slave rep"
+
+# LU-2844 mdt prepare fail should not cause umount oops
+test_111 ()
+{
+ [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.62) ]] ||
+ { skip "Need MDS version at least 2.3.62"; return 0; }
+
+ local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
+#define OBD_FAIL_MDS_CHANGELOG_INIT 0x151
+ do_facet $SINGLEMDS lctl set_param fail_loc=0x151
+ stop $SINGLEMDS || error "stop MDS failed"
+ start $SINGLEMDS $mdsdev && error "start MDS should fail"
+ do_facet $SINGLEMDS lctl set_param fail_loc=0
+ start $SINGLEMDS $mdsdev || error "start MDS failed"
+}
+run_test 111 "mdd setup fail should not cause umount oops"
+
+# LU-793
+test_112a() {
+ remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+
+ do_facet_random_file client $TMP/$tfile 100K ||
+ error_noexit "Create random file $TMP/$tfile"
+
+ pause_bulk "cp $TMP/$tfile $DIR/$tfile" $TIMEOUT ||
+ error_noexit "Can't pause_bulk copy"
+
+ df $DIR
+ # expect cmp to succeed, client resent bulk
+ cmp $TMP/$tfile $DIR/$tfile ||
+ error_noexit "Wrong data has been written"
+ rm $DIR/$tfile ||
+ error_noexit "Can't remove file"
+ rm $TMP/$tfile
+}
+run_test 112a "bulk resend while orignal request is in progress"
+
+complete $SECONDS
check_and_cleanup_lustre
exit_status
git://git.whamcloud.com / fs / lustre-release.git / blobdiff