#!/bin/bash
-# FIXME - there is no reason to use all of these different
-# return codes, espcially when most of them are mapped to something
-# else anyway. The combination of test number and return code
-# figure out what failed.
+# FIXME - there is no reason to use all of these different return codes,
+# espcially when most of them are mapped to something else anyway.
+# The tests should use error() to describe the failure more clearly,
+# and reduce the need to look into the tests to see what failed.
set -e
fi
[ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
-# bug number for skipped test: LU-2778 LU-2059
- ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h"
+# bug number for skipped test: LU-2778 LU-2059 LU-4444
+ ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h 69"
init_logging
require_dsh_mds || exit 0
require_dsh_ost || exit 0
#
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
assert_DIR
start_mgs () {
echo "start mgs"
- start mgs $MGSDEV $MGS_MOUNT_OPTS
+ start mgs $(mgsdevname) $MGS_MOUNT_OPTS
}
start_mdt() {
}
run_test 28 "permanent parameter setting"
+test_28a() { # LU-4221
+ [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
+ { skip "Need OST version at least 2.5.52" && return 0; }
+ [ "$(facet_fstype ost1)" = "zfs" ] &&
+ skip "LU-4221: no such proc params for ZFS OSTs" && return
+
+ local name
+ local param
+ local cmd
+ local old
+ local new
+ local device="$FSNAME-OST0000"
+
+ setup
+
+ # In this test we will set three kinds of proc parameters with
+ # lctl conf_param:
+ # 1. the ones moved from the OFD to the OSD, and only their
+ # symlinks kept in obdfilter
+ # 2. non-symlink ones in the OFD
+ # 3. non-symlink ones in the OSD
+
+ # Check 1.
+ # prepare a symlink parameter in the OFD
+ name="writethrough_cache_enable"
+ param="$device.ost.$name"
+ cmd="$LCTL get_param -n obdfilter.$device.$name"
+
+ # conf_param the symlink parameter in the OFD
+ old=$(do_facet ost1 $cmd)
+ new=$(((old + 1) % 2))
+ set_conf_param_and_check ost1 "$cmd" "$param" $new ||
+ error "lctl conf_param $device.ost.$param=$new failed"
+
+ # conf_param the target parameter in the OSD
+ param="$device.osd.$name"
+ cmd="$LCTL get_param -n osd-*.$device.$name"
+ set_conf_param_and_check ost1 "$cmd" "$param" $old ||
+ error "lctl conf_param $device.osd.$param=$old failed"
+
+ # Check 2.
+ # prepare a non-symlink parameter in the OFD
+ name="client_cache_seconds"
+ param="$device.ost.$name"
+ cmd="$LCTL get_param -n obdfilter.$device.$name"
+
+ # conf_param the parameter in the OFD
+ old=$(do_facet ost1 $cmd)
+ new=$((old * 2))
+ set_conf_param_and_check ost1 "$cmd" "$param" $new ||
+ error "lctl conf_param $device.ost.$param=$new failed"
+ set_conf_param_and_check ost1 "$cmd" "$param" $old ||
+ error "lctl conf_param $device.ost.$param=$old failed"
+
+ # Check 3.
+ # prepare a non-symlink parameter in the OSD
+ name="lma_self_repair"
+ param="$device.osd.$name"
+ cmd="$LCTL get_param -n osd-*.$device.$name"
+
+ # conf_param the parameter in the OSD
+ old=$(do_facet ost1 $cmd)
+ new=$(((old + 1) % 2))
+ set_conf_param_and_check ost1 "$cmd" "$param" $new ||
+ error "lctl conf_param $device.osd.$param=$new failed"
+ set_conf_param_and_check ost1 "$cmd" "$param" $old ||
+ error "lctl conf_param $device.osd.$param=$old failed"
+
+ cleanup
+}
+run_test 28a "set symlink parameters permanently with conf_param"
+
test_29() {
[ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
setup > /dev/null 2>&1
t32_wait_til_devices_gone() {
local node=$1
local devices
+ local loops
local i=0
echo wait for devices to go
while ((i < 20)); do
devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
- ((devices == 0)) && return 0
+ loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
+ ((devices == 0 && loops == 0)) && return 0
sleep 5
i=$((i + 1))
done
- echo "waiting for devices on $node: Given up"
+ echo "waiting for dev on $node: dev $devices loop $loops given up"
+ do_rpc_nodes $node "losetup -a"
+ do_rpc_nodes $node "$LCTL devices_list"
return 1
}
$r $LCTL set_param debug="$PTLDEBUG"
$r $TUNEFS --dryrun $tmp/mdt || {
+ $r losetup -a
error_noexit "tunefs.lustre before mounting the MDT"
return 1
}
mopts=loop,writeconf
if [ $fstype == "ldiskfs" ]; then
$r $TUNEFS --quota $tmp/mdt || {
+ $r losetup -a
error_noexit "Enable mdt quota feature"
return 1
}
$r mount -t lustre -o loop,nosvc $tmp/mdt $tmp/mnt/mdt
$r lctl replace_nids $fsname-OST0000 $ostnid
$r lctl replace_nids $fsname-MDT0000 $nid
- $r umount $tmp/mnt/mdt
+ $r umount -d $tmp/mnt/mdt
fi
mopts=loop,exclude=$fsname-OST0000
t32_wait_til_devices_gone $node
$r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || {
+ $r losetup -a
error_noexit "Mounting the MDT"
return 1
}
mopts=loop,mgsnode=$nid,$writeconf
if [ $fstype == "ldiskfs" ]; then
$r $TUNEFS --quota $tmp/ost || {
+ $r losetup -a
error_noexit "Enable ost quota feature"
return 1
}
# mount a second time to make sure we didnt leave upgrade flag on
$r $TUNEFS --dryrun $tmp/mdt || {
+ $r losetup -a
error_noexit "tunefs.lustre before remounting the MDT"
return 1
}
run_test 48 "too many acls on file"
# check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
-test_49() { # bug 17710
+test_49a() { # bug 17710
local timeout_orig=$TIMEOUT
local ldlm_timeout_orig=$LDLM_TIMEOUT
local LOCAL_TIMEOUT=20
reformat
setup_noconfig
- check_mount || return 1
+ check_mount || error "client mount failed"
echo "check ldlm_timout..."
- LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
- LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
- LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
+ local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
+ local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
+ local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
- if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
+ if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
fi
if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
- error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
+ error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
fi
umount_client $MOUNT
- stop_ost || return 2
- stop_mds || return 3
+ stop_ost || error "problem stopping OSS"
+ stop_mds || error "problem stopping MDS"
+
+ LDLM_TIMEOUT=$ldlm_timeout_orig
+ TIMEOUT=$timeout_orig
+}
+run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
+
+test_49b() { # bug 17710
+ local timeout_orig=$TIMEOUT
+ local ldlm_timeout_orig=$LDLM_TIMEOUT
+ local LOCAL_TIMEOUT=20
LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
+ TIMEOUT=$LOCAL_TIMEOUT
reformat
setup_noconfig
- check_mount || return 7
+ check_mount || error "client mount failed"
- LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
- LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
- LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
+ local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
+ local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
+ local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
- if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
+ if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
fi
if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
- error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
+ error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
fi
- cleanup || return $?
+ cleanup || error "cleanup failed"
LDLM_TIMEOUT=$ldlm_timeout_orig
TIMEOUT=$timeout_orig
}
-run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
+run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
lazystatfs() {
# Test both statfs and lfs df and fail if either one fails
# remove all files from the OBJECTS dir
do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
- do_facet $SINGLEMDS "umount $MNTDIR"
+ do_facet $SINGLEMDS "umount -d $MNTDIR"
# restart MDS with missing llog files
start_mds
do_facet mds "lctl set_param fail_loc=0"
do_facet $SINGLEMDS \
"mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
- do_facet $SINGLEMDS "umount $brpt"
+ do_facet $SINGLEMDS "umount -d $brpt"
# restart MDS, the "last_rcvd" file should be recreated.
start_mds || error "fail to restart the MDS"
run_test 68 "be able to reserve specific sequences in FLDB"
test_69() {
+ local server_version=$(lustre_version_code $SINGLEMDS)
+
+ [[ $server_version -lt $(version_code 2.4.2) ]] &&
+ skip "Need MDS version at least 2.4.2" && return
+
+ [[ $server_version -ge $(version_code 2.4.50) ]] &&
+ [[ $server_version -lt $(version_code 2.5.0) ]] &&
+ skip "Need MDS version at least 2.5.0" && return
+
setup
# use OST0000 since it probably has the most creations
# not, then the OST will refuse to allow the MDS connect
# because the LAST_ID value is too different from the MDS
#define OST_MAX_PRECREATE=20000
- local num_create=$((20000 * 3/2 - $last_id + 100))
+ local num_create=$((20000 * 5))
- mkdir $DIR/$tdir
+ mkdir -p $DIR/$tdir
$LFS setstripe -i 0 $DIR/$tdir
- createmany $DIR/$tdir/$tfile- $num_create
+ createmany -o $DIR/$tdir/$tfile- $num_create ||
+ error "createmany: failed to create $num_create files: $?"
# delete all of the files with objects on OST0 so the
# filesystem is not inconsistent later on
- $LFS find $MOUNT --index 0 -print0 | xargs -0 unlink
+ $LFS find $MOUNT --ost 0 | xargs rm
stop_ost || error "OST0 stop failure"
- add ost1 $(mkfs_opts ost1 $ostdev) --reformat --replace $ostdev ||
+ add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
+ $(ostdevname 1) $(ostvdevname 1) ||
error "reformat and replace $ostdev failed"
start_ost || error "OST0 restart failure"
+ wait_osc_import_state mds ost FULL
touch $DIR/$tdir/$tfile-last || error "create file after reformat"
- local idx=$($LFS getstripe -c $DIR/$tdir/$tfile-last)
+ local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
[ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
cleanup
}
-run_test 68 "replace an OST with the same index"
+run_test 69 "replace an OST with the same index"
test_70a() {
[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return