echo -n $size
}
+fs_inode_ksize() {
+ local facet=${1:-$SINGLEMDS}
+ local fstype=$(facet_fstype $facet)
+ local size=0
+ case $fstype in
+ ldiskfs) size=4;; # ~4KB per inode
+ zfs) size=11;; # 10 to 11KB per inode
+ esac
+
+ echo -n $size
+}
+
check_gss_daemon_nodes() {
local list=$1
dname=$2
echo -n $device
}
+running_in_vm() {
+ local virt=$(virt-what 2> /dev/null)
+
+ [ $? -eq 0 ] && [ -n "$virt" ] && { echo $virt; return; }
+
+ virt=$(dmidecode -s system-product-name | awk '{print $1}')
+
+ case $virt in
+ VMware|KVM|VirtualBox|Parallels) echo ${virt,,} ;;
+ *) ;;
+ esac
+}
+
#
# Re-read the partition table on failover partner host.
# After a ZFS storage pool is created on a shared device, the partition table
if [ -f /sys/devices/system/cpu/possible ]; then
local cpus=$(($(cut -d "-" -f 2 /sys/devices/system/cpu/possible)+1))
else
- local cpus=$(getconf _NPROCESSORS_CONF)
+ local cpus=$(getconf _NPROCESSORS_CONF 2>/dev/null)
fi
# bug 19944, adjust size to be -gt num_possible_cpus()
# commit the device label change to disk
if [[ $devicelabel =~ (:[a-zA-Z]{3}[0-9]{4}) ]]; then
+ echo "Commit the device label on ${!dev}"
do_facet $facet "sync; sync; sync"
+ sleep 5
fi
local client=$1
local mnt=$2
local force
- local busy
+ local busy
local need_kill
[ "$3" ] && force=-f
echo \\\$(hostname) env are INSANE!;
exit 1;
fi"
- [ $? -eq 0 ] || rc=1
+ [ $? -eq 0 ] || rc=1
done
return $rc
}
sanity_mount_check_servers () {
- [ "$CLIENTONLY" ] &&
+ [ -n "$CLIENTONLY" ] &&
{ echo "CLIENTONLY mode, skip mount_check_servers"; return 0; } || true
echo Checking servers environments
wait_for_function --quiet "! ping -w 3 -c 1 $host" 5 1 && return 0
echo "waiting for $host to fail attempts=$attempts"
[ $i -lt $attempts ] || \
- { echo "$host still pingable after power down! attempts=$attempts" && return 1; }
+ { echo "$host still pingable after power down! attempts=$attempts" && return 1; }
done
}
# only for remote client
check_client_load () {
- local client=$1
- local var=$(node_var_name $client)_load
- local TESTLOAD=run_${!var}.sh
-
- ps auxww | grep -v grep | grep $client | grep -q "$TESTLOAD" || return 1
-
- # bug 18914: try to connect several times not only when
- # check ps, but while check_catastrophe also
- local tries=3
- local RC=254
- while [ $RC = 254 -a $tries -gt 0 ]; do
- let tries=$tries-1
- # assume success
- RC=0
- if ! check_catastrophe $client; then
- RC=${PIPESTATUS[0]}
- if [ $RC -eq 254 ]; then
- # FIXME: not sure how long we shuold sleep here
- sleep 10
- continue
- fi
- echo "check catastrophe failed: RC=$RC "
- return $RC
- fi
- done
- # We can continue try to connect if RC=254
- # Just print the warning about this
- if [ $RC = 254 ]; then
- echo "got a return status of $RC from do_node while checking catastrophe on $client"
- fi
-
- # see if the load is still on the client
- tries=3
- RC=254
- while [ $RC = 254 -a $tries -gt 0 ]; do
- let tries=$tries-1
- # assume success
- RC=0
- if ! do_node $client "ps auxwww | grep -v grep | grep -q $TESTLOAD"; then
- RC=${PIPESTATUS[0]}
- sleep 30
- fi
- done
- if [ $RC = 254 ]; then
- echo "got a return status of $RC from do_node while checking (catastrophe and 'ps') the client load on $client"
- # see if we can diagnose a bit why this is
- fi
+ local client=$1
+ local var=$(node_var_name $client)_load
+ local testload=run_${!var}.sh
+
+ ps -C $testload | grep $client || return 1
+
+ # bug 18914: try to connect several times not only when
+ # check ps, but while check_node_health also
+
+ local tries=3
+ local RC=254
+ while [ $RC = 254 -a $tries -gt 0 ]; do
+ let tries=$tries-1
+ # assume success
+ RC=0
+ if ! check_node_health $client; then
+ RC=${PIPESTATUS[0]}
+ if [ $RC -eq 254 ]; then
+ # FIXME: not sure how long we shuold sleep here
+ sleep 10
+ continue
+ fi
+ echo "check node health failed: RC=$RC "
+ return $RC
+ fi
+ done
+ # We can continue try to connect if RC=254
+ # Just print the warning about this
+ if [ $RC = 254 ]; then
+ echo "got a return status of $RC from do_node while checking " \
+ "node health on $client"
+ fi
+
+ # see if the load is still on the client
+ tries=3
+ RC=254
+ while [ $RC = 254 -a $tries -gt 0 ]; do
+ let tries=$tries-1
+ # assume success
+ RC=0
+ if ! do_node $client \
+ "ps auxwww | grep -v grep | grep -q $testload"; then
+ RC=${PIPESTATUS[0]}
+ sleep 30
+ fi
+ done
+ if [ $RC = 254 ]; then
+ echo "got a return status of $RC from do_node while checking " \
+ "(node health and 'ps') the client load on $client"
+ # see if we can diagnose a bit why this is
+ fi
- return $RC
+ return $RC
}
check_client_loads () {
local clients=${1//,/ }
return $rc
}
+lfs_df_check() {
+ local clients=${1:-$CLIENTS}
+
+ if [ -z "$clients" ]; then
+ $LFS df $MOUNT
+ else
+ $PDSH $clients "$LFS df $MOUNT" > /dev/null
+ fi
+}
+
+
clients_up() {
- # not every config has many clients
- sleep 1
- if [ ! -z "$CLIENTS" ]; then
- $PDSH $CLIENTS "stat -f $MOUNT" > /dev/null
- else
- stat -f $MOUNT > /dev/null
- fi
+ # not every config has many clients
+ sleep 1
+ lfs_df_check
}
client_up() {
- local client=$1
- # usually checked on particular client or locally
- sleep 1
- if [ ! -z "$client" ]; then
- $PDSH $client "stat -f $MOUNT" > /dev/null
- else
- stat -f $MOUNT > /dev/null
- fi
+ # usually checked on particular client or locally
+ sleep 1
+ lfs_df_check $1
}
client_evicted() {
mount_ldiskfs() {
local facet=$1
local dev=$(facet_device $facet)
- local mnt=$(facet_mntpt $facet)
+ local mnt=${2:-$(facet_mntpt $facet)}
local opts
if ! do_facet $facet test -b $dev; then
unmount_ldiskfs() {
local facet=$1
local dev=$(facet_device $facet)
- local mnt=$(facet_mntpt $facet)
+ local mnt=${2:-$(facet_mntpt $facet)}
do_facet $facet $UMOUNT $mnt
}
mount_zfs() {
local facet=$1
local ds=$(facet_device $facet)
- local mnt=$(facet_mntpt $facet)
+ local mnt=${2:-$(facet_mntpt $facet)}
local canmnt
local mntpt
unmount_zfs() {
local facet=$1
local ds=$(facet_device $facet)
- local mnt=$(facet_mntpt $facet)
+ local mnt=${2:-$(facet_mntpt $facet)}
local var_mntpt=mz_$(var_name ${facet}_$ds)_mountpoint
local var_canmnt=mz_$(var_name ${facet}_$ds)_canmount
local mntpt=${!var_mntpt}
mount_fstype() {
local facet=$1
+ local mnt=$2
local fstype=$(facet_fstype $facet)
- mount_$fstype $facet
+ mount_$fstype $facet $mnt
}
unmount_fstype() {
local facet=$1
+ local mnt=$2
local fstype=$(facet_fstype $facet)
- unmount_$fstype $facet
+ unmount_$fstype $facet $mnt
}
########
zconf_umount_clients $clients $MOUNT "$*" || true
[ -n "$MOUNT2" ] && zconf_umount_clients $clients $MOUNT2 "$*" || true
- [ "$CLIENTONLY" ] && return
+ [ -n "$CLIENTONLY" ] && return
+
# The add fn does rm ${facet}active file, this would be enough
# if we use do_facet <facet> only after the facet added, but
# currently we use do_facet mds in local.sh
}
cleanup_echo_devs () {
- local devs=$($LCTL dl | grep echo | awk '{print $4}')
+ trap 0
+ local dev
+ local devs=$($LCTL dl | grep echo | awk '{print $4}')
- for dev in $devs; do
- $LCTL --device $dev cleanup
- $LCTL --device $dev detach
- done
+ for dev in $devs; do
+ $LCTL --device $dev cleanup
+ $LCTL --device $dev detach
+ done
}
cleanupall() {
formatall() {
stopall
# Set hostid for ZFS/SPL zpool import protection
- do_rpc_nodes "$(comma_list $(remote_nodes_list))" set_hostid
+ # (Assumes MDS version is also OSS version)
+ if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.8.54) ];
+ then
+ do_rpc_nodes "$(comma_list $(remote_nodes_list))" set_hostid
+ fi
# We need ldiskfs here, may as well load them all
load_modules
- [ "$CLIENTONLY" ] && return
+ [ -n "$CLIENTONLY" ] && return
echo Formatting mgs, mds, osts
if ! combined_mgs_mds ; then
format_mgs
}
setupall() {
- nfs_client_mode && return
+ nfs_client_mode && return
cifs_client_mode && return
- sanity_mount_check ||
- error "environments are insane!"
+ sanity_mount_check || error "environments are insane!"
- load_modules
+ load_modules
- if [ -z "$CLIENTONLY" ]; then
- echo Setup mgs, mdt, osts
- echo $WRITECONF | grep -q "writeconf" && \
- writeconf_all
- if ! combined_mgs_mds ; then
+ if [ -z "$CLIENTONLY" ]; then
+ echo Setup mgs, mdt, osts
+ echo $WRITECONF | grep -q "writeconf" && writeconf_all
+ if ! combined_mgs_mds ; then
start mgs $(mgsdevname) $MGS_MOUNT_OPTS
- fi
+ fi
for num in `seq $MDSCOUNT`; do
DEVNAME=$(mdsdevname $num)
}
init_facet_vars () {
- [ "$CLIENTONLY" ] && return 0
+ [ -n "$CLIENTONLY" ] && return 0
local facet=$1
shift
local device=$1
}
init_param_vars () {
- remote_mds_nodsh ||
- TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
+ TIMEOUT=$(lctl get_param -n timeout)
+ TIMEOUT=${TIMEOUT:-20}
+ remote_mds_nodsh && log "Using TIMEOUT=$TIMEOUT" && return 0
+
+ TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
log "Using TIMEOUT=$TIMEOUT"
osc_ensure_active $SINGLEMDS $TIMEOUT
local mntpt=$1
local mounted=$(mount | grep " $mntpt ")
- if [ "$CLIENTONLY" ]; then
+ if [ -n "$CLIENTONLY" ]; then
# bug 18021
# CLIENTONLY should not depend on *_HOST settings
local mgc=$($LCTL device_list | awk '/MGC/ {print $4}')
# in theory someone could create a new,
# client-only config file that assumed lustre was already
# configured and didn't set the MGSNID. If MGSNID is not set,
- # then we should use the mgs nid currently being used
+ # then we should use the mgs nid currently being used
# as the default value. bug 18021
[[ x$MGSNID = x ]] &&
MGSNID=${mgc//MGC/}
return 0
fi
- local myMGS_host=$mgs_HOST
+ local myMGS_host=$mgs_HOST
if [ "$NETTYPE" = "ptl" ]; then
- myMGS_host=$(h2ptl $mgs_HOST | sed -e s/@ptl//)
+ myMGS_host=$(h2ptl $mgs_HOST | sed -e s/@ptl//)
fi
echo Checking config lustre mounted on $mntpt
restore_mount $MOUNT2
export I_MOUNTED2=yes
fi
- fi
+ fi
# 5.
# MOUNT is mounted MOUNT2 is not mounted
set_default_debug_nodes $(comma_list $(nodes_list))
fi
- if [ $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ] ; then
+ if [ -z "$CLIENTONLY" -a $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ]; then
local facets=""
[ "$(facet_fstype ost1)" = "ldiskfs" ] &&
facets="$(get_facets OST)"
set_flavor_all $SEC
fi
- #Enable remote MDT create for testing
- for num in $(seq $MDSCOUNT); do
- do_facet mds$num \
- lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
- 2>/dev/null
- done
+ if [ -z "$CLIENTONLY" ]; then
+ # Enable remote MDT create for testing
+ for num in $(seq $MDSCOUNT); do
+ do_facet mds$num \
+ lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
+ 2>/dev/null
+ done
+ fi
if [ "$ONLY" == "setup" ]; then
exit 0
true
DEBUGSAVE=""
- [ -n "DEBUGSAVE_SERVER" ] &&
+ [ -n "$DEBUGSAVE_SERVER" ] &&
do_nodes $(comma_list $(all_server_nodes)) \
"$LCTL set_param debug=\\\"${DEBUGSAVE_SERVER}\\\"" ||
true
exit 1
}
+# Throw an error if it's not running in vm - usually for performance
+# verification
+error_not_in_vm() {
+ local virt=$(running_in_vm)
+ if [[ -n "$virt" ]]; then
+ echo "running in VM '$virt', ignore error"
+ error_ignore env=$virt "$@"
+ else
+ error "$@"
+ fi
+}
+
skip_env () {
$FAIL_ON_SKIP_ENV && error false $@ || skip $@
}
#
-# Log a message (on all nodes) padded with "=" before and after.
+# Log a message (on all nodes) padded with "=" before and after.
# Also appends a timestamp and prepends the testsuite name.
-#
+#
EQUALS="===================================================================================================="
banner() {
cd $SAVE_PWD
reset_fail_loc
check_grant ${testnum} || error "check_grant $testnum failed with $?"
- check_catastrophe || error "LBUG/LASSERT detected"
+ check_node_health
check_dmesg_for_errors || error "Error in dmesg detected"
if [ "$PARALLEL" != "yes" ]; then
ps auxww | grep -v grep | grep -q multiop &&
remote_mds_nodsh()
{
- [ "$CLIENTONLY" ] && return 0 || true
- remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
+ [ -n "$CLIENTONLY" ] && return 0 || true
+ remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
}
require_dsh_mds()
{
- remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" && \
- MSKIPPED=1 && return 1
- return 0
+ remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" &&
+ MSKIPPED=1 && return 1
+ return 0
}
remote_ost ()
remote_ost_nodsh()
{
- [ "$CLIENTONLY" ] && return 0 || true
- remote_ost && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
+ [ -n "$CLIENTONLY" ] && return 0 || true
+ remote_ost && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
}
require_dsh_ost()
remote_mgs_nodsh()
{
- [ "$CLIENTONLY" ] && return 0 || true
- local MGS
- MGS=$(facet_host mgs)
- remote_node $MGS && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
+ [ -n "$CLIENTONLY" ] && return 0 || true
+ local MGS
+ MGS=$(facet_host mgs)
+ remote_node $MGS && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
}
local_mode ()
rnodes=${rnodes//,/ }
local -a nodes=($rnodes)
- local num=${#nodes[@]}
+ local num=${#nodes[@]}
local i=$((RANDOM * num * 2 / 65536))
echo ${nodes[i]}
}
client_only () {
- [ "$CLIENTONLY" ] || [ "$CLIENTMODSONLY" = yes ]
+ [ -n "$CLIENTONLY" ] || [ "x$CLIENTMODSONLY" = "xyes" ]
}
check_versions () {
done
}
-check_catastrophe() {
+check_node_health() {
local nodes=${1:-$(comma_list $(nodes_list))}
- do_nodes $nodes "rc=0;
-val=\\\$($LCTL get_param -n catastrophe 2>&1);
-if [[ \\\$? -eq 0 && \\\$val -ne 0 ]]; then
- echo \\\$(hostname -s): \\\$val;
- rc=\\\$val;
-fi;
-exit \\\$rc"
+ for node in ${nodes//,/ }; do
+ check_network "$node" 5
+ if [ $? -eq 0 ]; then
+ do_node $node "rc=0;
+ val=\\\$($LCTL get_param -n catastrophe 2>&1);
+ if [[ \\\$? -eq 0 && \\\$val -ne 0 ]]; then
+ echo \\\$(hostname -s): \\\$val;
+ rc=\\\$val;
+ fi;
+ exit \\\$rc" || error "$node:LBUG/LASSERT detected"
+ fi
+ done
}
mdsrate_cleanup () {
suffix="$ts.log"
echo "Dumping lctl log to ${prefix}.*.${suffix}"
- if [ "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
+ if [ -n "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
echo "Dumping logs only on local client."
$LCTL dk > ${prefix}.debug_log.$(hostname -s).${suffix}
dmesg > ${prefix}.dmesg.$(hostname -s).${suffix}
#
get_page_size() {
local facet=$1
- local size
+ local size=$(getconf PAGE_SIZE 2>/dev/null)
- size=$(do_facet $facet getconf PAGE_SIZE)
- [[ ${PIPESTATUS[0]} = 0 && -n "$size" ]] || size=4096
- echo -n $size
+ [ -z "$CLIENTONLY" ] && size=$(do_facet $facet getconf PAGE_SIZE)
+ echo -n ${size:-4096}
}
#
local device=$2
local count
- count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+ [ -z "$CLIENTONLY" ] && count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
awk '/^Block count:/ {print $3}')
- echo -n $count
+ echo -n ${count:-0}
}
# Get the block size of the filesystem.
get_block_size() {
- local facet=$1
- local device=$2
- local size
+ local facet=$1
+ local device=$2
+ local size
- size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
- awk '/^Block size:/ {print $3}')
- echo $size
+ [ -z "$CLIENTONLY" ] && size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+ awk '/^Block size:/ {print $3}')
+ echo -n ${size:-0}
}
# Check whether the "large_xattr" feature is enabled or not.
git://git.whamcloud.com / fs / lustre-release.git / blobdiff