[ ! -f "$LCTL" ] && export LCTL=$(which lctl)
export LFS=${LFS:-"$LUSTRE/utils/lfs"}
[ ! -f "$LFS" ] && export LFS=$(which lfs)
+ export KSOCKLND_CONFIG=${KSOCKLND_CONFIG:-"$LUSTRE/scripts/ksocklnd-config"}
+ [ ! -f "$KSOCKLND_CONFIG" ] &&
+ export KSOCKLND_CONFIG=$(which ksocklnd-config 2> /dev/null)
export PERM_CMD=${PERM_CMD:-"$LCTL conf_param"}
/sbin/lsmod | grep -q "^\<$1\>"
}
+check_lfs_df_ret_val() {
+ # Ignore only EOPNOTSUPP (which is 95; Operation not supported) error
+ # returned by 'lfs df' for valid dentry but not a lustrefs.
+ #
+ # 'lfs df' historically always returned success(0) instead of
+ # EOPNOTSUPP. This function for compatibility reason, ignores and
+ # masquerades EOPNOTSUPP as success.
+ [[ $1 -eq 95 ]] && return 0
+ return $1
+}
+
PRLFS=false
lustre_insmod() {
local module=$1
# that obviously has nothing to do with this Lustre run
# Disable automatic memory scanning to avoid perf hit.
if [ -f /sys/kernel/debug/kmemleak ] ; then
- echo scan=off > /sys/kernel/debug/kmemleak
- echo scan > /sys/kernel/debug/kmemleak
- echo clear > /sys/kernel/debug/kmemleak
+ echo scan=off > /sys/kernel/debug/kmemleak || true
+ echo scan > /sys/kernel/debug/kmemleak || true
+ echo clear > /sys/kernel/debug/kmemleak || true
fi
echo Loading modules from $LUSTRE
local size=0
case $(facet_fstype $facet) in
- ldiskfs) size=50;; # largest seen is 44, leave some headroom
+ ldiskfs) size=72;; # largest seen is 64, leave some headroom
# grant_block_size is in bytes, allow at least 2x max blocksize
zfs) size=$(lctl get_param osc.$FSNAME*.import |
awk '/grant_block_size:/ {print $2/512; exit;}')
local device=${2:-$FSNAME-OST*}
local name=$3
- do_nodes $nodes "$LCTL get_param -n obdfilter.$device.$name \
- osd-*.$device.$name 2>&1" | grep -v 'error:'
+ do_nodes $nodes "$LCTL get_param -n osd-*.$device.$name"
}
set_osd_param() {
local name=$3
local value=$4
- do_nodes $nodes "$LCTL set_param -n obdfilter.$device.$name=$value \
- osd-*.$device.$name=$value 2>&1" | grep -v 'error:'
+ do_nodes $nodes "$LCTL set_param -n osd-*.$device.$name=$value"
}
set_debug_size () {
if [ "$old_MDT_QUOTA_TYPE" ]; then
if [[ $PERM_CMD == *"set_param -P"* ]]; then
do_facet mgs $PERM_CMD \
- osd-*.$FSNAME-MDT*.quota_slave.enable = \
+ osd-*.$FSNAME-MDT*.quota_slave.enabled = \
$old_MDT_QUOTA_TYPE
else
do_facet mgs $PERM_CMD \
if [ "$old_OST_QUOTA_TYPE" ]; then
if [[ $PERM_CMD == *"set_param -P"* ]]; then
do_facet mgs $PERM_CMD \
- osd-*.$FSNAME-OST*.quota_slave.enable = \
+ osd-*.$FSNAME-OST*.quota_slave.enabled = \
$old_OST_QUOTA_TYPE
else
do_facet mgs $LCTL conf_param \
# This will allow fixing the "lfs df" summary line in the future.
lfs_df() {
$LFS df $* | sed -e 's/filesystem /filesystem_/'
+ check_lfs_df_ret_val $?
}
# Get free inodes on the MDT specified by mdt index, free indoes on
if [[ $PERM_CMD == *"set_param -P"* ]]; then
do_facet mgs $PERM_CMD \
- osd-*.$FSNAME-MDT*.quota_slave.enable=$QUOTA_TYPE
+ osd-*.$FSNAME-MDT*.quota_slave.enabled=$QUOTA_TYPE
do_facet mgs $PERM_CMD \
- osd-*.$FSNAME-OST*.quota_slave.enable=$QUOTA_TYPE
+ osd-*.$FSNAME-OST*.quota_slave.enabled=$QUOTA_TYPE
else
do_facet mgs $PERM_CMD $FSNAME.quota.mdt=$QUOTA_TYPE ||
error "set mdt quota type failed"
reboot_facet() {
local facet=$1
local node=$(facet_active_host $facet)
+ local sleep_time=${2:-10}
if [ "$FAILURE_MODE" = HARD ]; then
boot_node $node
else
- sleep 10
+ sleep $sleep_time
fi
}
# If --verbose is passed as the first argument, the result is printed on each
# value change, otherwise it is only printed after every 10s interval.
#
+# If --quiet is passed as the first/second argument, the do_node() command
+# will not print the remote command before executing it each time.
+#
# Using wait_update_cond() or related helper function is preferable to adding
# a "long enough" wait for some state to change in the background, since
# "long enough" may be too short due to tunables, system config, or running in
# a VM, and must by necessity wait too long for most cases or risk failure.
#
-# usage: wait_update_cond [--verbose] node check cond expect [max_wait]
+# usage: wait_update_cond [--verbose] [--quiet] node check cond expect [max_wait]
wait_update_cond() {
- local verbose=false
- [[ "$1" == "--verbose" ]] && verbose=true && shift
+ local verbose
+ local quiet
+
+ [[ "$1" == "--verbose" ]] && verbose="$1" && shift
+ [[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
local node=$1
local check="$2"
local print=10
while (( $waited <= $max_wait )); do
- result=$(do_node $node "$check")
+ result=$(do_node $quiet $node "$check")
eval [[ "'$result'" $cond "'$expect'" ]]
if [[ $? == 0 ]]; then
echo "Updated after ${waited}s: want '$expect' got '$result'"
return 0
fi
- if $verbose && [[ "$result" != "$prev_result" ]]; then
+ if [[ -n "$verbose" && "$result" != "$prev_result" ]]; then
[[ -n "$prev_result" ]] &&
echo "Changed after ${waited}s: from '$prev_result' to '$result'"
prev_result="$result"
return 3
}
-# usage: wait_update [--verbose] node check expect [max_wait]
+# usage: wait_update [--verbose] [--quiet] node check expect [max_wait]
wait_update() {
- local verbose=
- [ "$1" = "--verbose" ] && verbose="$1" && shift
+ local verbose
+ local quiet
+
+ [[ "$1" == "--verbose" ]] && verbose="$1" && shift
+ [[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
local node="$1"
local check="$2"
local expect="$3"
local max_wait=$4
- wait_update_cond $verbose $node "$check" "==" "$expect" $max_wait
+ wait_update_cond $verbose $quiet $node "$check" "==" "$expect" $max_wait
}
# usage: wait_update_facet_cond [--verbose] facet check cond expect [max_wait]
wait_update_facet_cond() {
- local verbose=
- [ "$1" = "--verbose" ] && verbose="$1" && shift
+ local verbose
+ local quiet
+
+ [[ "$1" == "--verbose" ]] && verbose="$1" && shift
+ [[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
local node=$(facet_active_host $1)
local check="$2"
local expect="$4"
local max_wait=$5
- wait_update_cond $verbose $node "$check" "$cond" "$expect" $max_wait
+ wait_update_cond $verbose $quiet $node "$check" "$cond" "$expect" $max_wait
}
# usage: wait_update_facet [--verbose] facet check expect [max_wait]
wait_update_facet() {
- local verbose=
- [ "$1" = "--verbose" ] && verbose="$1" && shift
+ local verbose
+ local quiet
+
+ [[ "$1" == "--verbose" ]] && verbose="$1" && shift
+ [[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
local node=$(facet_active_host $1)
local check="$2"
local expect="$3"
local max_wait=$4
- wait_update_cond $verbose $node "$check" "==" "$expect" $max_wait
+ wait_update_cond $verbose $quiet $node "$check" "==" "$expect" $max_wait
}
sync_all_data() {
# MAX value shouldn't be big as this mean server responsiveness
# never increase this just to make test pass but investigate
# why it takes so long time
- local MAX=5
+ local MAX=${1:-5}
local WAIT=0
local list=$(comma_list $(mdts_nodes))
while [ $WAIT -lt $MAX ]; do
lfs_df_check() {
local clients=${1:-$CLIENTS}
+ local rc
if [ -z "$clients" ]; then
- $LFS df $MOUNT
+ $LFS df $MOUNT > /dev/null
+ rc=$?
else
$PDSH $clients "$LFS df $MOUNT" > /dev/null
+ rc=$?
fi
+
+ check_lfs_df_ret_val $rc
+ rc=$?
+
+ return $rc
}
clients_up() {
lfs_df_check
}
+all_mds_up() {
+ (( MDSCOUNT == 1 )) && return
+
+ # wait so that statfs data on MDT expire
+ local delay=$(do_facet $SINGLEMDS lctl \
+ get_param -n osp.*MDT0000*MDT0001.maxage)
+ sleep $delay
+ local nodes=$(comma_list $(mdts_nodes))
+ # initiate statfs RPC, all to all MDTs
+ do_nodes $nodes $LCTL get_param -N osp.*MDT*MDT*.filesfree >&/dev/null
+ do_nodes $nodes $LCTL get_param -N osp.*MDT*MDT*.filesfree >&/dev/null
+}
+
client_up() {
# usually checked on particular client or locally
sleep 1
facet=$(echo ${affecteds[index]} | tr -s " " | cut -d"," -f 1)
echo reboot facets: ${affecteds[index]}
- reboot_facet $facet
+ reboot_facet $facet $sleep_time
change_active ${affecteds[index]}
mount_facet $facet -o $abort_type
clients_up || echo "first stat failed: $?"
clients_up || error "post-failover stat: $?"
+ all_mds_up
}
host_nids_address() {
}
do_node() {
- local verbose=false
- # do not stripe off hostname if verbose, bug 19215
- if [ x$1 = x--verbose ]; then
- shift
- verbose=true
- fi
+ local verbose
+ local quiet
- local HOST=$1
- shift
- local myPDSH=$PDSH
- if [ "$HOST" = "$HOSTNAME" ]; then
- myPDSH="no_dsh"
- elif [ -z "$myPDSH" -o "$myPDSH" = "no_dsh" ]; then
- echo "cannot run remote command on $HOST with $myPDSH"
- return 128
- fi
- if $VERBOSE; then
- echo "CMD: $HOST $@" >&2
- $myPDSH $HOST "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
- fi
+ # do not strip off hostname if verbose, b=19215
+ [[ "$1" == "--verbose" ]] && verbose="$1" && shift
+ [[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
+
+ local HOST=$1
+ shift
+ local myPDSH=$PDSH
+
+ if [ "$HOST" = "$HOSTNAME" ]; then
+ myPDSH="no_dsh"
+ elif [ -z "$myPDSH" -o "$myPDSH" = "no_dsh" ]; then
+ echo "cannot run remote command on $HOST with $myPDSH"
+ return 128
+ fi
+ if $VERBOSE && [[ -z "$quiet" ]]; then
+ echo "CMD: $HOST $@" >&2
+ $myPDSH $HOST "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
+ fi
if [[ "$myPDSH" == "rsh" ]] ||
[[ "$myPDSH" == *pdsh* && "$myPDSH" != *-S* ]]; then
return 0
fi
- if $verbose ; then
+ if [[ -n "$verbose" ]]; then
# print HOSTNAME for myPDSH="no_dsh"
if [[ $myPDSH = no_dsh ]]; then
$myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" sh -c \"$@\")" | sed -e "s/^/${HOSTNAME}: /"
return ${PIPESTATUS[0]}
}
-do_nodev() {
- do_node --verbose "$@"
-}
-
single_local_node () {
- [ "$1" = "$HOSTNAME" ]
+ [ "$1" = "$HOSTNAME" ]
}
# Outputs environment variable assignments that should be passed to remote nodes
}
do_nodes() {
- local verbose=false
- # do not stripe off hostname if verbose, bug 19215
- if [ x$1 = x--verbose ]; then
- shift
- verbose=true
- fi
+ local verbose
+ local quiet
- local rnodes=$1
- shift
+ # do not strip off hostname if verbose, b=19215
+ [[ "$1" == "--verbose" ]] && verbose="$1" && shift
+ [[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
- if single_local_node $rnodes; then
- if $verbose; then
- do_nodev $rnodes "$@"
- else
- do_node $rnodes "$@"
- fi
- return $?
- fi
+ local rnodes=$1
+ shift
- # This is part from do_node
- local myPDSH=$PDSH
+ if single_local_node $rnodes; then
+ do_node $verbose $quiet $rnodes "$@"
+ return $?
+ fi
- [ -z "$myPDSH" -o "$myPDSH" = "no_dsh" -o "$myPDSH" = "rsh" ] && \
- echo "cannot run remote command on $rnodes with $myPDSH" && return 128
+ # This is part from do_node
+ local myPDSH=$PDSH
- export FANOUT=$(get_node_count "${rnodes//,/ }")
- if $VERBOSE; then
- echo "CMD: $rnodes $@" >&2
- $myPDSH $rnodes "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
- fi
+ [ -z "$myPDSH" -o "$myPDSH" = "no_dsh" -o "$myPDSH" = "rsh" ] &&
+ echo "cannot run remote command on $rnodes with $myPDSH" &&
+ return 128
- # do not replace anything from pdsh output if -N is used
- # -N Disable hostname: prefix on lines of output.
- if $verbose || [[ $myPDSH = *-N* ]]; then
- $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) sh -c \"$@\")"
- else
- $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) sh -c \"$@\")" | sed -re "s/^[^:]*: //g"
- fi
- return ${PIPESTATUS[0]}
+ export FANOUT=$(get_node_count "${rnodes//,/ }")
+ if $VERBOSE && [[ -z "$quiet" ]]; then
+ echo "CMD: $rnodes $@" >&2
+ $myPDSH $rnodes "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
+ fi
+
+ # do not replace anything from pdsh output if -N is used
+ # -N Disable hostname: prefix on lines of output.
+ if [[ -n "$verbose" || $myPDSH = *-N* ]]; then
+ $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) sh -c \"$@\")"
+ else
+ $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) sh -c \"$@\")" | sed -re "s/^[^:]*: //g"
+ fi
+ return ${PIPESTATUS[0]}
}
##
#
# usage: do_facet $facet command [arg ...]
do_facet() {
+ local verbose
+ local quiet
+
+ [[ "$1" == "--verbose" ]] && verbose="$1" && shift
+ [[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
+
local facet=$1
shift
- local HOST=$(facet_active_host $facet)
- [ -z $HOST ] && echo "No host defined for facet ${facet}" && exit 1
- do_node $HOST "$@"
+ local host=$(facet_active_host $facet)
+
+ [ -z "$host" ] && echo "No host defined for facet ${facet}" && exit 1
+ do_node $verbose $quiet $host "$@"
}
# Function: do_facet_random_file $FACET $FILE $SIZE
}
do_nodesv() {
- do_nodes --verbose "$@"
+ do_nodes --verbose "$@"
}
add() {
log "Using TIMEOUT=$TIMEOUT"
# tune down to speed up testing on (usually) small setups
+ local mgc_timeout=/sys/module/mgc/parameters/mgc_requeue_timeout_min
do_nodes $(comma_list $(nodes_list)) \
- "echo 1 >/sys/module/mgc/parameters/mgc_requeue_timeout_min"
+ "[ -f $mgc_timeout ] && echo 1 > $mgc_timeout; exit 0"
osc_ensure_active $SINGLEMDS $TIMEOUT
osc_ensure_active client $TIMEOUT
(( MDS1_VERSION <= $(version_code 2.13.52) )) ||
do_nodes $(comma_list $(mdts_nodes)) \
"$LCTL set_param lod.*.mdt_hash=crush"
+ do_node $(mgs_node) "$LCTL set_param -P *.*.lbug_on_grant_miscount=1"
return 0
}
done
}
+set_pools_quota () {
+ local u
+ local o
+ local p
+ local i
+ local j
+
+ [[ $ENABLE_QUOTA ]] || error "Required Pool Quotas: \
+ $POOLS_QUOTA_USERS_SET, but ENABLE_QUOTA not set!"
+
+ # POOLS_QUOTA_USERS_SET=
+ # "quota15_1:20M -- for all of the found pools
+ # quota15_2:1G:gpool0
+ # quota15_3 -- for global limit only
+ # quota15_4:200M:gpool0
+ # quota15_4:200M:gpool1"
+
+ declare -a pq_userset=(${POOLS_QUOTA_USERS_SET="mpiuser"})
+ declare -a pq_users
+ declare -A pq_limits
+
+ for ((i=0; i<${#pq_userset[@]}; i++)); do
+ u=${pq_userset[i]%%:*}
+ o=""
+ # user gets no pool limits if
+ # POOLS_QUOTA_USERS_SET does not specify it
+ [[ ${pq_userset[i]} =~ : ]] && o=${pq_userset[i]##$u:}
+ pq_limits[$u]+=" $o"
+ done
+ pq_users=(${!pq_limits[@]})
+
+ declare -a opts
+ local pool
+
+ for ((i=0; i<${#pq_users[@]}; i++)); do
+ u=${pq_users[i]}
+ # set to max limit (_u64)
+ $LFS setquota -u $u -B $((2**24 - 1))T $DIR
+ opts=(${pq_limits[$u]})
+ for ((j=0; j<${#opts[@]}; j++)); do
+ p=${opts[j]##*:}
+ o=${opts[j]%%:*}
+ # Set limit for all existing pools if
+ # no pool specified
+ if [ $p == $o ]; then
+ p=$(list_pool $FSNAME | sed "s/$FSNAME.//")
+ echo "No pool specified for $u,
+ set limit $o for all existing pools"
+ fi
+ for pool in $p; do
+ $LFS setquota -u $u -B $o --pool $pool $DIR ||
+ error "setquota -u $u -B $o \
+ --pool $pool failed"
+ done
+ done
+ $LFS quota -uv $u --pool $DIR
+ done
+}
+
check_and_setup_lustre() {
sanitize_parameters
nfs_client_mode && return
create_pools $FS_POOL $FS_POOL_NOSTS
fi
+ if [[ -n "$POOLS_QUOTA_USERS_SET" ]]; then
+ set_pools_quota
+ fi
if [ "$ONLY" == "setup" ]; then
exit 0
fi
# General functions
wait_for_function () {
- local quiet=""
-
- # suppress fn both stderr and stdout
- if [ "$1" = "--quiet" ]; then
- shift
- quiet=" > /dev/null 2>&1"
+ local quiet=""
- fi
+ # suppress fn both stderr and stdout
+ if [ "$1" = "--quiet" ]; then
+ shift
+ quiet=" > /dev/null 2>&1"
+ fi
- local fn=$1
- local max=${2:-900}
- local sleep=${3:-5}
+ local fn=$1
+ local max=${2:-900}
+ local sleep=${3:-5}
- local wait=0
+ local wait=0
- while true; do
+ while true; do
- eval $fn $quiet && return 0
+ eval $fn $quiet && return 0
- wait=$((wait + sleep))
- [ $wait -lt $max ] || return 1
- echo waiting $fn, $((max - wait)) secs left ...
- sleep $sleep
- done
+ [ $wait -lt $max ] || return 1
+ echo waiting $fn, $((max - wait)) secs left ...
+ wait=$((wait + sleep))
+ [ $wait -gt $max ] && ((sleep -= wait - max))
+ sleep $sleep
+ done
}
check_network() {
}
reset_fail_loc () {
- echo -n "Resetting fail_loc on all nodes..."
- do_nodes $(comma_list $(nodes_list)) "lctl set_param -n fail_loc=0 \
- fail_val=0 2>/dev/null" || true
- echo done.
+ #echo -n "Resetting fail_loc on all nodes..."
+ do_nodes --quiet $(comma_list $(nodes_list)) \
+ "lctl set_param -n fail_loc=0 fail_val=0 2>/dev/null" || true
+ #echo done.
}
# Also appends a timestamp and prepends the testsuite name.
#
-EQUALS="===================================================================================================="
+# ======================================================== 15:06:12 (1624050372)
+EQUALS="========================================================"
banner() {
msg="== ${TESTSUITE} $*"
last=${msg: -1:1}
ldiskfs_check_descriptors: Checksum for group 0 failed\|\
group descriptors corrupted"
- res=$(do_nodes $(comma_list $(nodes_list)) "dmesg" | grep "$errors")
+ res=$(do_nodes -q $(comma_list $(nodes_list)) "dmesg" | grep "$errors")
[ -z "$res" ] && return 0
echo "Kernel error detected: $res"
return 1
# sync all the data and make sure no pending data on server
do_nodes $clients sync
- clients_up # initiate all idling connections
+ do_nodes $clients $LFS df # initiate all idling connections
# get client grant
cli_grant=$(grant_from_clients $clients)
}
ostname_from_index() {
- local uuid=$(ostuuid_from_index $1)
+ local uuid=$(ostuuid_from_index $1 $2)
echo ${uuid/_UUID/}
}
check_node_health() {
local nodes=${1:-$(comma_list $(nodes_list))}
-
- for node in ${nodes//,/ }; do
- check_network "$node" 5
- if [ $? -eq 0 ]; then
- do_node $node "$LCTL get_param catastrophe 2>&1" |
- grep -q "catastrophe=1" &&
- error "$node:LBUG/LASSERT detected" || true
- fi
- done
+ local health=$TMP/node_health.$$
+
+ do_nodes -q $nodes "$LCTL get_param catastrophe 2>&1" | tee $health |
+ grep "catastrophe=1" && error "LBUG/LASSERT detected"
+ # Only check/report network health if get_param isn't reported, since
+ # *clearly* the network is working if get_param returned something.
+ if (( $(grep -c catastro $health) != $(wc -w <<< ${nodes//,/ }) )); then
+ for node in ${nodes//,/}; do
+ check_network $node 5
+ done
+ fi
+ rm -f $health
}
mdsrate_cleanup () {
}
_wait_import_state () {
- local expected=$1
- local CONN_PROC=$2
- local maxtime=${3:-$(max_recovery_time)}
- local error_on_failure=${4:-1}
- local CONN_STATE
- local i=0
+ local expected="$1"
+ local CONN_PROC="$2"
+ local maxtime=${3:-$(max_recovery_time)}
+ local err_on_fail=${4:-1}
+ local CONN_STATE
+ local i=0
CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
- while ! echo "${CONN_STATE}" | egrep -q "^${expected}\$" ; do
- if [ "${expected}" == "DISCONN" ]; then
- # for disconn we can check after proc entry is removed
- [ "x${CONN_STATE}" == "x" ] && return 0
- # with AT enabled, we can have connect request timeout near of
- # reconnect timeout and test can't see real disconnect
- [ "${CONN_STATE}" == "CONNECTING" ] && return 0
- fi
- if [ $i -ge $maxtime ]; then
- [ $error_on_failure -ne 0 ] && \
- error "can't put import for $CONN_PROC into ${expected}" \
- "state after $i sec, have ${CONN_STATE}"
- return 1
- fi
- sleep 1
- # Add uniq for multi-mount case
- CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
- i=$(($i + 1))
- done
+ while ! echo "${CONN_STATE}" | egrep -q "^${expected}\$" ; do
+ if [[ "${expected}" == "DISCONN" ]]; then
+ # for disconn we can check after proc entry is removed
+ [[ -z "${CONN_STATE}" ]] && return 0
+ # with AT, we can have connect request timeout near
+ # reconnect timeout and test can't see real disconnect
+ [[ "${CONN_STATE}" == "CONNECTING" ]] && return 0
+ fi
+ if (( $i >= $maxtime )); then
+ (( $err_on_fail != 0 )) &&
+ error "can't put import for $CONN_PROC into ${expected} state after $i sec, have ${CONN_STATE}"
+ return 1
+ fi
+ sleep 1
+ # Add uniq for multi-mount case
+ CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null |
+ cut -f2 | uniq)
+ i=$((i + 1))
+ done
- log "$CONN_PROC in ${CONN_STATE} state after $i sec"
- return 0
+ log "$CONN_PROC in ${CONN_STATE} state after $i sec"
+ return 0
}
wait_import_state() {
- local state=$1
- local params=$2
- local maxtime=${3:-$(max_recovery_time)}
- local error_on_failure=${4:-1}
- local param
-
- for param in ${params//,/ }; do
- _wait_import_state $state $param $maxtime $error_on_failure || return
- done
+ local expected="$1"
+ local params="$2"
+ local maxtime=${3:-$(max_recovery_time)}
+ local err_on_fail=${4:-1}
+ local param
+
+ for param in ${params//,/ }; do
+ _wait_import_state "$expected" "$param" $maxtime $err_on_fail ||
+ return
+ done
}
wait_import_state_mount() {
return 0
fi
- wait_import_state $*
+ wait_import_state "$@"
}
# One client request could be timed out because server was not ready
}
wait_clients_import_state () {
- local list=$1
- local facet=$2
- local expected=$3
-
- local facets=$facet
+ local list="$1"
+ local facet="$2"
+ local expected="$3"
+ local facets="$facet"
if [ "$FAILURE_MODE" = HARD ]; then
facets=$(facets_on_host $(facet_active_host $facet))
local proc_path
case $facet in
ost* ) proc_path="osc.$(get_clientosc_proc_path \
- $label).ost_server_uuid" ;;
+ $label).ost_server_uuid" ;;
mds* ) proc_path="mdc.$(get_clientmdc_proc_path \
- $label).mds_server_uuid" ;;
+ $label).mds_server_uuid" ;;
mgs* ) proc_path="mgc.$(get_clientmgc_proc_path \
- $label).mgs_server_uuid" ;;
+ $label).mgs_server_uuid" ;;
*) error "unknown facet!" ;;
esac
error "$mdt: changelog_mask=+hsm failed: $?"
local cl_user
- cl_user=$(do_facet $facet \
- $LCTL --device $mdt changelog_register -n) ||
+ cl_user=$(do_facet $facet $LCTL --device $mdt \
+ changelog_register -n $@) ||
error "$mdt: register changelog user failed: $?"
stack_trap "__changelog_deregister $facet $cl_user" EXIT
}
# LU-13417: XXX lots of tests assume the directory to be created under MDT0,
-# using this function to create directory under MDT0 explicitly.
-# Don't use it in new tests, and remove it from old tests.
+# created on MDT0, use this function to create directory on specific MDT
+# explicitly, and set default LMV to create subdirs on the same MDT too.
+mkdir_on_mdt() {
+ local mdt
+ local OPTIND=1
+
+ while getopts "i:" opt $*; do
+ case $opt in
+ i) mdt=$OPTARG;;
+ esac
+ done
+
+ shift $((OPTIND - 1))
+
+ $LFS mkdir -i $mdt -c 1 $*
+ # setting default LMV in non-DNE system will cause sanity-quota 41 fail
+ ((MDSCOUNT < 2)) || $LFS setdirstripe -D -i $mdt -c 1 $*
+}
+
mkdir_on_mdt0() {
- $LFS mkdir -i 0 -c 1 $*
+ mkdir_on_mdt -i0 $*
+}
+
+# Wait for nodemap synchronization
+wait_nm_sync() {
+ local nodemap_name=$1
+ local key=$2
+ local value=$3
+ local opt=$4
+ local proc_param
+ local is_active=$(do_facet mgs $LCTL get_param -n nodemap.active)
+ local max_retries=20
+ local is_sync
+ local out1=""
+ local out2
+ local mgs_ip=$(host_nids_address $mgs_HOST $NETTYPE | cut -d' ' -f1)
+ local i
+
+ if [ "$nodemap_name" == "active" ]; then
+ proc_param="active"
+ elif [ -z "$key" ]; then
+ proc_param=${nodemap_name}
+ else
+ proc_param="${nodemap_name}.${key}"
+ fi
+ if [ "$opt" == "inactive" ]; then
+ # check nm sync even if nodemap is not activated
+ is_active=1
+ opt=""
+ fi
+ (( is_active == 0 )) && [ "$proc_param" != "active" ] && return
+
+ if [ -z "$value" ]; then
+ out1=$(do_facet mgs $LCTL get_param $opt \
+ nodemap.${proc_param} 2>/dev/null)
+ echo "On MGS ${mgs_ip}, ${proc_param} = $out1"
+ else
+ out1=$value;
+ fi
+
+ # if servers run on the same node, it is impossible to tell if they get
+ # synced with the mgs, so just wait an arbitrary 10 seconds
+ if [ $(facet_active_host mgs) == $(facet_active_host mds) ] &&
+ [ $(facet_active_host mgs) == $(facet_active_host ost1) ]; then
+ echo "waiting 10 secs for sync"
+ sleep 10
+ return
+ fi
+
+ # wait up to 10 seconds for other servers to sync with mgs
+ for i in $(seq 1 10); do
+ for node in $(all_server_nodes); do
+ local node_ip=$(host_nids_address $node $NETTYPE |
+ cut -d' ' -f1)
+
+ is_sync=true
+ if [ -z "$value" ]; then
+ [ $node_ip == $mgs_ip ] && continue
+ fi
+
+ out2=$(do_node $node_ip $LCTL get_param $opt \
+ nodemap.$proc_param 2>/dev/null)
+ echo "On $node ${node_ip}, ${proc_param} = $out2"
+ [ "$out1" != "$out2" ] && is_sync=false && break
+ done
+ $is_sync && break
+ sleep 1
+ done
+ if ! $is_sync; then
+ echo MGS
+ echo $out1
+ echo OTHER - IP: $node_ip
+ echo $out2
+ error "mgs and $nodemap_name ${key} mismatch, $i attempts"
+ fi
+ echo "waited $((i - 1)) seconds for sync"
+}
+
+consume_precreations() {
+ local dir=$1
+ local mfacet=$2
+ local OSTIDX=$3
+ local extra=${4:-2}
+ local OST=$(ostname_from_index $OSTIDX $dir)
+
+ test_mkdir -p $dir/${OST}
+ $LFS setstripe -i $OSTIDX -c 1 ${dir}/${OST}
+
+ # on the mdt's osc
+ local mdtosc_proc=$(get_mdtosc_proc_path $mfacet $OST)
+ local last_id=$(do_facet $mfacet $LCTL get_param -n \
+ osp.$mdtosc_proc.prealloc_last_id)
+ local next_id=$(do_facet $mfacet $LCTL get_param -n \
+ osp.$mdtosc_proc.prealloc_next_id)
+ echo "Creating to objid $last_id on ost $OST..."
+ createmany -o $dir/${OST}/f $next_id $((last_id - next_id + extra))
+}
+
+__exhaust_precreations() {
+ local OSTIDX=$1
+ local FAILLOC=$2
+ local FAILIDX=${3:-$OSTIDX}
+ local ofacet=ost$((OSTIDX + 1))
+
+ mkdir_on_mdt0 $DIR/$tdir
+ local mdtidx=$($LFS getstripe -m $DIR/$tdir)
+ local mfacet=mds$((mdtidx + 1))
+ echo OSTIDX=$OSTIDX MDTIDX=$mdtidx
+
+ local mdtosc_proc=$(get_mdtosc_proc_path $mfacet)
+ do_facet $mfacet $LCTL get_param osp.$mdtosc_proc.prealloc*
+
+#define OBD_FAIL_OST_ENOSPC 0x215
+ do_facet $ofacet $LCTL set_param fail_val=$FAILIDX fail_loc=0x215
+
+ consume_precreations $DIR/$tdir $mfacet $OSTIDX
+
+ do_facet $mfacet $LCTL get_param osp.$mdtosc_proc.prealloc*
+ do_facet $ofacet $LCTL set_param fail_loc=$FAILLOC
+}
+
+exhaust_precreations() {
+ __exhaust_precreations $1 $2 $3
+ sleep_maxage
+}
+
+exhaust_all_precreations() {
+ local i
+ for (( i=0; i < OSTCOUNT; i++ )) ; do
+ __exhaust_precreations $i $1 -1
+ done
+ sleep_maxage
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff