#!/bin/bash
-trap 'print_summary && touch $TF_FAIL && \
+trap 'print_summary && print_stack_trace | tee $TF_FAIL && \
echo "$TESTSUITE: FAIL: test-framework exiting on error"' ERR
set -e
-#set -x
export LANG=en_US
export REFORMAT=${REFORMAT:-""}
export WRITECONF=${WRITECONF:-""}
export VERBOSE=${VERBOSE:-false}
-export GSS=false
+export GSS=${GSS:-false}
+export GSS_SK=${GSS_SK:-false}
export GSS_KRB5=false
export GSS_PIPEFS=false
+export SHARED_KEY=${SHARED_KEY:-false}
+export SK_PATH=${SK_PATH:-/tmp/test-framework-keys}
+export SK_OM_PATH=$SK_PATH'/tmp-request-mount'
+export SK_MOUNTED=${SK_MOUNTED:-false}
+export SK_FLAVOR=${SK_FLAVOR:-ski}
+export SK_NO_KEY=${SK_NO_KEY:-true}
+export SK_UNIQUE_NM=${SK_UNIQUE_NM:-false}
+export SK_S2S=${SK_S2S:-false}
+export SK_S2SNM=${SK_S2SNM:-TestFrameNM}
+export SK_S2SNMCLI=${SK_S2SNMCLI:-TestFrameNMCli}
export IDENTITY_UPCALL=default
export QUOTA_AUTO=1
# specify environment variable containing batch job name for server statistics
export DO_CLEANUP=${DO_CLEANUP:-true}
export KEEP_ZPOOL=${KEEP_ZPOOL:-false}
- export MKE2FS=$MKE2FS
- if [ -z "$MKE2FS" ]; then
- if which mkfs.ldiskfs >/dev/null 2>&1; then
- export MKE2FS=mkfs.ldiskfs
- else
- export MKE2FS=mke2fs
- fi
- fi
+ export MKE2FS=$MKE2FS
+ if [ -z "$MKE2FS" ]; then
+ if which mkfs.ldiskfs >/dev/null 2>&1; then
+ export MKE2FS=mkfs.ldiskfs
+ else
+ export MKE2FS=mke2fs
+ fi
+ fi
- export DEBUGFS=$DEBUGFS
- if [ -z "$DEBUGFS" ]; then
- if which debugfs.ldiskfs >/dev/null 2>&1; then
- export DEBUGFS=debugfs.ldiskfs
- else
- export DEBUGFS=debugfs
- fi
- fi
+ export DEBUGFS=$DEBUGFS
+ if [ -z "$DEBUGFS" ]; then
+ if which debugfs.ldiskfs >/dev/null 2>&1; then
+ export DEBUGFS=debugfs.ldiskfs
+ else
+ export DEBUGFS=debugfs
+ fi
+ fi
- export TUNE2FS=$TUNE2FS
- if [ -z "$TUNE2FS" ]; then
- if which tunefs.ldiskfs >/dev/null 2>&1; then
- export TUNE2FS=tunefs.ldiskfs
- else
- export TUNE2FS=tune2fs
- fi
- fi
+ export TUNE2FS=$TUNE2FS
+ if [ -z "$TUNE2FS" ]; then
+ if which tunefs.ldiskfs >/dev/null 2>&1; then
+ export TUNE2FS=tunefs.ldiskfs
+ else
+ export TUNE2FS=tune2fs
+ fi
+ fi
- export E2LABEL=$E2LABEL
- if [ -z "$E2LABEL" ]; then
- if which label.ldiskfs >/dev/null 2>&1; then
- export E2LABEL=label.ldiskfs
- else
- export E2LABEL=e2label
- fi
- fi
+ export E2LABEL=$E2LABEL
+ if [ -z "$E2LABEL" ]; then
+ if which label.ldiskfs >/dev/null 2>&1; then
+ export E2LABEL=label.ldiskfs
+ else
+ export E2LABEL=e2label
+ fi
+ fi
- export DUMPE2FS=$DUMPE2FS
- if [ -z "$DUMPE2FS" ]; then
- if which dumpfs.ldiskfs >/dev/null 2>&1; then
- export DUMPE2FS=dumpfs.ldiskfs
- else
- export DUMPE2FS=dumpe2fs
- fi
- fi
+ export DUMPE2FS=$DUMPE2FS
+ if [ -z "$DUMPE2FS" ]; then
+ if which dumpfs.ldiskfs >/dev/null 2>&1; then
+ export DUMPE2FS=dumpfs.ldiskfs
+ else
+ export DUMPE2FS=dumpe2fs
+ fi
+ fi
- export E2FSCK=$E2FSCK
- if [ -z "$E2FSCK" ]; then
- if which fsck.ldiskfs >/dev/null 2>&1; then
- export E2FSCK=fsck.ldiskfs
- else
- export E2FSCK=e2fsck
- fi
- fi
+ export E2FSCK=$E2FSCK
+ if [ -z "$E2FSCK" ]; then
+ if which fsck.ldiskfs >/dev/null 2>&1; then
+ export E2FSCK=fsck.ldiskfs
+ else
+ export E2FSCK=e2fsck
+ fi
+ fi
export RESIZE2FS=$RESIZE2FS
if [ -z "$RESIZE2FS" ]; then
export ZDB=${ZDB:-zdb}
export PARTPROBE=${PARTPROBE:-partprobe}
- #[ -d /r ] && export ROOT=${ROOT:-/r}
- export TMP=${TMP:-$ROOT/tmp}
- export TESTSUITELOG=${TMP}/${TESTSUITE}.log
- export LOGDIR=${LOGDIR:-${TMP}/test_logs/$(date +%s)}
- export TESTLOG_PREFIX=$LOGDIR/$TESTSUITE
+ #[ -d /r ] && export ROOT=${ROOT:-/r}
+ export TMP=${TMP:-$ROOT/tmp}
+ export TESTSUITELOG=${TMP}/${TESTSUITE}.log
+ export LOGDIR=${LOGDIR:-${TMP}/test_logs/$(date +%s)}
+ export TESTLOG_PREFIX=$LOGDIR/$TESTSUITE
- export HOSTNAME=${HOSTNAME:-$(hostname -s)}
- if ! echo $PATH | grep -q $LUSTRE/utils; then
- export PATH=$LUSTRE/utils:$PATH
- fi
- if ! echo $PATH | grep -q $LUSTRE/utils/gss; then
- export PATH=$LUSTRE/utils/gss:$PATH
- fi
- if ! echo $PATH | grep -q $LUSTRE/tests; then
- export PATH=$LUSTRE/tests:$PATH
- fi
- if ! echo $PATH | grep -q $LUSTRE/../lustre-iokit/sgpdd-survey; then
- export PATH=$LUSTRE/../lustre-iokit/sgpdd-survey:$PATH
- fi
- export LST=${LST:-"$LUSTRE/../lnet/utils/lst"}
- [ ! -f "$LST" ] && export LST=$(which lst)
- export SGPDDSURVEY=${SGPDDSURVEY:-"$LUSTRE/../lustre-iokit/sgpdd-survey/sgpdd-survey")}
- [ ! -f "$SGPDDSURVEY" ] && export SGPDDSURVEY=$(which sgpdd-survey)
+ export HOSTNAME=${HOSTNAME:-$(hostname -s)}
+ if ! echo $PATH | grep -q $LUSTRE/utils; then
+ export PATH=$LUSTRE/utils:$PATH
+ fi
+ if ! echo $PATH | grep -q $LUSTRE/utils/gss; then
+ export PATH=$LUSTRE/utils/gss:$PATH
+ fi
+ if ! echo $PATH | grep -q $LUSTRE/tests; then
+ export PATH=$LUSTRE/tests:$PATH
+ fi
+ if ! echo $PATH | grep -q $LUSTRE/../lustre-iokit/sgpdd-survey; then
+ export PATH=$LUSTRE/../lustre-iokit/sgpdd-survey:$PATH
+ fi
+ export LST=${LST:-"$LUSTRE/../lnet/utils/lst"}
+ [ ! -f "$LST" ] && export LST=$(which lst)
+ export SGPDDSURVEY=${SGPDDSURVEY:-"$LUSTRE/../lustre-iokit/sgpdd-survey/sgpdd-survey")}
+ [ ! -f "$SGPDDSURVEY" ] && export SGPDDSURVEY=$(which sgpdd-survey)
export MCREATE=${MCREATE:-mcreate}
- # Ubuntu, at least, has a truncate command in /usr/bin
- # so fully path our truncate command.
- export TRUNCATE=${TRUNCATE:-$LUSTRE/tests/truncate}
+ export MULTIOP=${MULTIOP:-multiop}
+ # Ubuntu, at least, has a truncate command in /usr/bin
+ # so fully path our truncate command.
+ export TRUNCATE=${TRUNCATE:-$LUSTRE/tests/truncate}
export FSX=${FSX:-$LUSTRE/tests/fsx}
- export MDSRATE=${MDSRATE:-"$LUSTRE/tests/mpi/mdsrate"}
- [ ! -f "$MDSRATE" ] && export MDSRATE=$(which mdsrate 2> /dev/null)
- if ! echo $PATH | grep -q $LUSTRE/tests/racer; then
- export PATH=$LUSTRE/tests/racer:$PATH:
- fi
- if ! echo $PATH | grep -q $LUSTRE/tests/mpi; then
- export PATH=$LUSTRE/tests/mpi:$PATH
- fi
- export RSYNC_RSH=${RSYNC_RSH:-rsh}
-
- export LCTL=${LCTL:-"$LUSTRE/utils/lctl"}
- [ ! -f "$LCTL" ] && export LCTL=$(which lctl)
- export LFS=${LFS:-"$LUSTRE/utils/lfs"}
- [ ! -f "$LFS" ] && export LFS=$(which lfs)
- SETSTRIPE=${SETSTRIPE:-"$LFS setstripe"}
- GETSTRIPE=${GETSTRIPE:-"$LFS getstripe"}
-
- export L_GETIDENTITY=${L_GETIDENTITY:-"$LUSTRE/utils/l_getidentity"}
- if [ ! -f "$L_GETIDENTITY" ]; then
- if `which l_getidentity > /dev/null 2>&1`; then
- export L_GETIDENTITY=$(which l_getidentity)
- else
- export L_GETIDENTITY=NONE
- fi
- fi
- export LL_DECODE_FILTER_FID=${LL_DECODE_FILTER_FID:-"$LUSTRE/utils/ll_decode_filter_fid"}
- [ ! -f "$LL_DECODE_FILTER_FID" ] && export LL_DECODE_FILTER_FID="ll_decode_filter_fid"
- export LL_DECODE_LINKEA=${LL_DECODE_LINKEA:-"$LUSTRE/utils/ll_decode_linkea"}
- [ ! -f "$LL_DECODE_LINKEA" ] && export LL_DECODE_LINKEA="ll_decode_linkea"
- export MKFS=${MKFS:-"$LUSTRE/utils/mkfs.lustre"}
- [ ! -f "$MKFS" ] && export MKFS="mkfs.lustre"
- export TUNEFS=${TUNEFS:-"$LUSTRE/utils/tunefs.lustre"}
- [ ! -f "$TUNEFS" ] && export TUNEFS="tunefs.lustre"
- export CHECKSTAT="${CHECKSTAT:-"checkstat -v"} "
- export LUSTRE_RMMOD=${LUSTRE_RMMOD:-$LUSTRE/scripts/lustre_rmmod}
- [ ! -f "$LUSTRE_RMMOD" ] &&
- export LUSTRE_RMMOD=$(which lustre_rmmod 2> /dev/null)
- export LFS_MIGRATE=${LFS_MIGRATE:-$LUSTRE/scripts/lfs_migrate}
- [ ! -f "$LFS_MIGRATE" ] &&
- export LFS_MIGRATE=$(which lfs_migrate 2> /dev/null)
- export LR_READER=${LR_READER:-"$LUSTRE/utils/lr_reader"}
- [ ! -f "$LR_READER" ] && export LR_READER=$(which lr_reader 2> /dev/null)
- [ -z "$LR_READER" ] && export LR_READER="/usr/sbin/lr_reader"
- export NAME=${NAME:-local}
- export LGSSD=${LGSSD:-"$LUSTRE/utils/gss/lgssd"}
- [ "$GSS_PIPEFS" = "true" ] && [ ! -f "$LGSSD" ] && \
- export LGSSD=$(which lgssd)
- export LSVCGSSD=${LSVCGSSD:-"$LUSTRE/utils/gss/lsvcgssd"}
- [ ! -f "$LSVCGSSD" ] && export LSVCGSSD=$(which lsvcgssd 2> /dev/null)
- export KRB5DIR=${KRB5DIR:-"/usr/kerberos"}
- export DIR2
- export SAVE_PWD=${SAVE_PWD:-$LUSTRE/tests}
- export AT_MAX_PATH
- export LDEV=${LDEV:-"$LUSTRE/scripts/ldev"}
- [ ! -f "$LDEV" ] && export LDEV=$(which ldev 2> /dev/null)
-
- if [ "$ACCEPTOR_PORT" ]; then
- export PORT_OPT="--port $ACCEPTOR_PORT"
- fi
+ export MDSRATE=${MDSRATE:-"$LUSTRE/tests/mpi/mdsrate"}
+ [ ! -f "$MDSRATE" ] && export MDSRATE=$(which mdsrate 2> /dev/null)
+ if ! echo $PATH | grep -q $LUSTRE/tests/racer; then
+ export PATH=$LUSTRE/tests/racer:$PATH:
+ fi
+ if ! echo $PATH | grep -q $LUSTRE/tests/mpi; then
+ export PATH=$LUSTRE/tests/mpi:$PATH
+ fi
+ export RSYNC_RSH=${RSYNC_RSH:-rsh}
+
+ export LCTL=${LCTL:-"$LUSTRE/utils/lctl"}
+ [ ! -f "$LCTL" ] && export LCTL=$(which lctl)
+ export LFS=${LFS:-"$LUSTRE/utils/lfs"}
+ [ ! -f "$LFS" ] && export LFS=$(which lfs)
+ SETSTRIPE=${SETSTRIPE:-"$LFS setstripe"}
+ GETSTRIPE=${GETSTRIPE:-"$LFS getstripe"}
+
+ export L_GETIDENTITY=${L_GETIDENTITY:-"$LUSTRE/utils/l_getidentity"}
+ if [ ! -f "$L_GETIDENTITY" ]; then
+ if `which l_getidentity > /dev/null 2>&1`; then
+ export L_GETIDENTITY=$(which l_getidentity)
+ else
+ export L_GETIDENTITY=NONE
+ fi
+ fi
+ export LL_DECODE_FILTER_FID=${LL_DECODE_FILTER_FID:-"$LUSTRE/utils/ll_decode_filter_fid"}
+ [ ! -f "$LL_DECODE_FILTER_FID" ] && export LL_DECODE_FILTER_FID="ll_decode_filter_fid"
+ export LL_DECODE_LINKEA=${LL_DECODE_LINKEA:-"$LUSTRE/utils/ll_decode_linkea"}
+ [ ! -f "$LL_DECODE_LINKEA" ] && export LL_DECODE_LINKEA="ll_decode_linkea"
+ export MKFS=${MKFS:-"$LUSTRE/utils/mkfs.lustre"}
+ [ ! -f "$MKFS" ] && export MKFS="mkfs.lustre"
+ export TUNEFS=${TUNEFS:-"$LUSTRE/utils/tunefs.lustre"}
+ [ ! -f "$TUNEFS" ] && export TUNEFS="tunefs.lustre"
+ export CHECKSTAT="${CHECKSTAT:-"checkstat -v"} "
+ export LUSTRE_RMMOD=${LUSTRE_RMMOD:-$LUSTRE/scripts/lustre_rmmod}
+ [ ! -f "$LUSTRE_RMMOD" ] &&
+ export LUSTRE_RMMOD=$(which lustre_rmmod 2> /dev/null)
+ export LFS_MIGRATE=${LFS_MIGRATE:-$LUSTRE/scripts/lfs_migrate}
+ [ ! -f "$LFS_MIGRATE" ] &&
+ export LFS_MIGRATE=$(which lfs_migrate 2> /dev/null)
+ export LR_READER=${LR_READER:-"$LUSTRE/utils/lr_reader"}
+ [ ! -f "$LR_READER" ] &&
+ export LR_READER=$(which lr_reader 2> /dev/null)
+ [ -z "$LR_READER" ] && export LR_READER="/usr/sbin/lr_reader"
+ export NAME=${NAME:-local}
+ export LGSSD=${LGSSD:-"$LUSTRE/utils/gss/lgssd"}
+ [ "$GSS_PIPEFS" = "true" ] && [ ! -f "$LGSSD" ] &&
+ export LGSSD=$(which lgssd)
+ export LSVCGSSD=${LSVCGSSD:-"$LUSTRE/utils/gss/lsvcgssd"}
+ [ ! -f "$LSVCGSSD" ] && export LSVCGSSD=$(which lsvcgssd 2> /dev/null)
+ export KRB5DIR=${KRB5DIR:-"/usr/kerberos"}
+ export DIR2
+ export SAVE_PWD=${SAVE_PWD:-$LUSTRE/tests}
+ export AT_MAX_PATH
+ export LDEV=${LDEV:-"$LUSTRE/scripts/ldev"}
+ [ ! -f "$LDEV" ] && export LDEV=$(which ldev 2> /dev/null)
+
+ if [ "$ACCEPTOR_PORT" ]; then
+ export PORT_OPT="--port $ACCEPTOR_PORT"
+ fi
+
+ if $SHARED_KEY; then
+ $RPC_MODE || echo "Using GSS shared-key feature"
+ which lgss_sk > /dev/null 2>&1 ||
+ error_exit "built with lgss_sk disabled! SEC=$SEC"
+ GSS=true
+ GSS_SK=true
+ SEC=$SK_FLAVOR
+ fi
+
+ case "x$SEC" in
+ xkrb5*)
+ $RPC_MODE || echo "Using GSS/krb5 ptlrpc security flavor"
+ which lgss_keyring > /dev/null 2>&1 ||
+ error_exit "built with gss disabled! SEC=$SEC"
+ GSS=true
+ GSS_KRB5=true
+ ;;
+ esac
+
+ case "x$IDUP" in
+ xtrue)
+ IDENTITY_UPCALL=true
+ ;;
+ xfalse)
+ IDENTITY_UPCALL=false
+ ;;
+ esac
+
+ export LOAD_MODULES_REMOTE=${LOAD_MODULES_REMOTE:-false}
- case "x$SEC" in
- xkrb5*)
- echo "Using GSS/krb5 ptlrpc security flavor"
- which lgss_keyring > /dev/null 2>&1 || \
- error_exit "built with gss disabled! SEC=$SEC"
- GSS=true
- GSS_KRB5=true
- ;;
- esac
-
- case "x$IDUP" in
- xtrue)
- IDENTITY_UPCALL=true
- ;;
- xfalse)
- IDENTITY_UPCALL=false
- ;;
- esac
-
- export LOAD_MODULES_REMOTE=${LOAD_MODULES_REMOTE:-false}
-
- # Paths on remote nodes, if different
- export RLUSTRE=${RLUSTRE:-$LUSTRE}
- export RPWD=${RPWD:-$PWD}
- export I_MOUNTED=${I_MOUNTED:-"no"}
+ # Paths on remote nodes, if different
+ export RLUSTRE=${RLUSTRE:-$LUSTRE}
+ export RPWD=${RPWD:-$PWD}
+ export I_MOUNTED=${I_MOUNTED:-"no"}
if [ ! -f /lib/modules/$(uname -r)/kernel/fs/lustre/mdt.ko -a \
! -f /lib/modules/$(uname -r)/updates/kernel/fs/lustre/mdt.ko -a \
! -f /lib/modules/$(uname -r)/extra/kernel/fs/lustre/mdt.ko -a \
# output: prints version string to stdout in (up to 4) dotted-decimal values
lustre_build_version() {
local facet=${1:-client}
- local ver
-
local ver=$(do_facet $facet "$LCTL get_param -n version 2>/dev/null ||
$LCTL lustre_build_version 2>/dev/null ||
$LCTL --version 2>/dev/null | cut -d' ' -f2")
- local lver=$(egrep -i "lustre: |version: " <<<$ver | head -n 1)
+ local lver=$(egrep -i "lustre: |version: " <<<"$ver" | head -n 1)
[ -n "$lver" ] && ver="$lver"
- sed -e 's/.*: //' -e 's/^v//' -e 's/-.*//' -e 's/_/./g' <<<$ver |
+ sed -e 's/[^:]*: //' -e 's/^v//' -e 's/[ -].*//' -e 's/_/./g' <<<$ver |
cut -d. -f1-4
}
set_default_debug
load_module ../lnet/lnet/lnet
- case $NETTYPE in
- o2ib)
- LNETLND="o2iblnd/ko2iblnd"
- ;;
- *)
- ;;
- esac
- LNETLND=${LNETLND:-"socklnd/ksocklnd"}
- load_module ../lnet/klnds/$LNETLND
- load_module obdclass/obdclass
- load_module ptlrpc/ptlrpc
- load_module ptlrpc/gss/ptlrpc_gss
- load_module fld/fld
- load_module fid/fid
- load_module lmv/lmv
- load_module mdc/mdc
- load_module osc/osc
- load_module lov/lov
- load_module mgc/mgc
- load_module obdecho/obdecho
+
+ LNDPATH=${LNDPATH:-"../lnet/klnds"}
+ if [ -z "$LNETLND" ]; then
+ case $NETTYPE in
+ o2ib*) LNETLND="o2iblnd/ko2iblnd" ;;
+ tcp*) LNETLND="socklnd/ksocklnd" ;;
+ *) local lnd="${NETTYPE%%[0-9]}lnd"
+ [ -f "$LNDPATH/$lnd/k$lnd.ko" ] &&
+ LNETLND="$lnd/k$lnd" ||
+ LNETLND="socklnd/ksocklnd"
+ esac
+ fi
+ load_module ../lnet/klnds/$LNETLND
+ load_module obdclass/obdclass
+ load_module ptlrpc/ptlrpc
+ load_module ptlrpc/gss/ptlrpc_gss
+ load_module fld/fld
+ load_module fid/fid
+ load_module lmv/lmv
+ load_module osc/osc
+ load_module mdc/mdc
+ load_module lov/lov
+ load_module mgc/mgc
+ load_module obdecho/obdecho
if ! client_only; then
SYMLIST=/proc/kallsyms
grep -q crc16 $SYMLIST ||
local size=0
case $fstype in
ldiskfs) size=50;; # largest seen is 44, leave some headroom
- zfs) size=400;; # largest seen is 384
+ zfs) size=512;; # largest seen is 512
esac
echo -n $size
do_nodes $list "killall -2 $@ 2>/dev/null || true"
}
-# start gss daemons on all nodes, or
-# "daemon" on "list" if set
+# start gss daemons on all nodes, or "daemon" on "nodes" if set
start_gss_daemons() {
- local list=$1
- local daemon=$2
+ local nodes=$1
+ local daemon=$2
- if [ "$list" ] && [ "$daemon" ] ; then
- echo "Starting gss daemon on nodes: $list"
- do_nodes $list "$daemon" || return 8
- return 0
- fi
+ if [ "$nodes" ] && [ "$daemon" ] ; then
+ echo "Starting gss daemon on nodes: $nodes"
+ do_nodes $nodes "$daemon" || return 8
+ return 0
+ fi
- local list=$(comma_list $(mdts_nodes))
- echo "Starting gss daemon on mds: $list"
- do_nodes $list "$LSVCGSSD -v" || return 1
- if $GSS_PIPEFS; then
- do_nodes $list "$LGSSD -v" || return 2
- fi
+ nodes=$(comma_list $(mdts_nodes))
+ echo "Starting gss daemon on mds: $nodes"
+ if $GSS_SK; then
+ # Start all versions, in case of switching
+ do_nodes $nodes "$LSVCGSSD -vvv -s -m -o -z" || return 1
+ else
+ do_nodes $nodes "$LSVCGSSD -v" || return 1
+ fi
+ if $GSS_PIPEFS; then
+ do_nodes $nodes "$LGSSD -v" || return 2
+ fi
- list=$(comma_list $(osts_nodes))
- echo "Starting gss daemon on ost: $list"
- do_nodes $list "$LSVCGSSD -v" || return 3
- # starting on clients
+ nodes=$(comma_list $(osts_nodes))
+ echo "Starting gss daemon on ost: $nodes"
+ if $GSS_SK; then
+ # Start all versions, in case of switching
+ do_nodes $nodes "$LSVCGSSD -vvv -s -m -o -z" || return 3
+ else
+ do_nodes $nodes "$LSVCGSSD -v" || return 3
+ fi
+ # starting on clients
- local clients=${CLIENTS:-`hostname`}
- if $GSS_PIPEFS; then
- echo "Starting $LGSSD on clients $clients "
- do_nodes $clients "$LGSSD -v" || return 4
- fi
+ local clients=${CLIENTS:-$HOSTNAME}
+ if $GSS_PIPEFS; then
+ echo "Starting $LGSSD on clients $clients "
+ do_nodes $clients "$LGSSD -v" || return 4
+ fi
- # wait daemons entering "stable" status
- sleep 5
-
- #
- # check daemons are running
- #
- list=$(comma_list $(mdts_nodes) $(osts_nodes))
- check_gss_daemon_nodes $list lsvcgssd || return 5
- if $GSS_PIPEFS; then
- list=$(comma_list $(mdts_nodes))
- check_gss_daemon_nodes $list lgssd || return 6
- fi
- if $GSS_PIPEFS; then
- check_gss_daemon_nodes $clients lgssd || return 7
- fi
+ # wait daemons entering "stable" status
+ sleep 5
+
+ #
+ # check daemons are running
+ #
+ nodes=$(comma_list $(mdts_nodes) $(osts_nodes))
+ check_gss_daemon_nodes $nodes lsvcgssd || return 5
+ if $GSS_PIPEFS; then
+ nodes=$(comma_list $(mdts_nodes))
+ check_gss_daemon_nodes $nodes lgssd || return 6
+ fi
+ if $GSS_PIPEFS; then
+ check_gss_daemon_nodes $clients lgssd || return 7
+ fi
}
stop_gss_daemons() {
- local list=$(comma_list $(mdts_nodes))
+ local nodes=$(comma_list $(mdts_nodes))
- send_sigint $list lsvcgssd lgssd
+ send_sigint $nodes lsvcgssd lgssd
- list=$(comma_list $(osts_nodes))
- send_sigint $list lsvcgssd
+ nodes=$(comma_list $(osts_nodes))
+ send_sigint $nodes lsvcgssd
- list=${CLIENTS:-`hostname`}
- send_sigint $list lgssd
+ nodes=${CLIENTS:-$HOSTNAME}
+ send_sigint $nodes lgssd
+}
+
+add_sk_mntflag() {
+ # Add mount flags for shared key
+ local mt_opts=$@
+ if grep -q skpath <<< "$mt_opts" ; then
+ mt_opts=$(echo $mt_opts |
+ sed -e "s#skpath=[^ ,]*#skpath=$SK_PATH#")
+ else
+ if [ -z "$mt_opts" ]; then
+ mt_opts="-o skpath=$SK_PATH"
+ else
+ mt_opts="$mt_opts,skpath=$SK_PATH"
+ fi
+ fi
+ echo -n $mt_opts
}
init_gss() {
- if $GSS; then
- if ! module_loaded ptlrpc_gss; then
- load_module ptlrpc/gss/ptlrpc_gss
- module_loaded ptlrpc_gss ||
- error_exit "init_gss : GSS=$GSS, but gss/krb5 is not supported!"
- fi
- if $GSS_KRB5; then
- start_gss_daemons || error_exit "start gss daemon failed! rc=$?"
- fi
+ if $SHARED_KEY; then
+ GSS=true
+ GSS_SK=true
+ fi
+
+ if ! $GSS; then
+ return
+ fi
+
+ if ! module_loaded ptlrpc_gss; then
+ load_module ptlrpc/gss/ptlrpc_gss
+ module_loaded ptlrpc_gss ||
+ error_exit "init_gss: GSS=$GSS, but gss/krb5 missing"
+ fi
+
+ if $GSS_KRB5 || $GSS_SK; then
+ start_gss_daemons || error_exit "start gss daemon failed! rc=$?"
+ fi
+
+ if $GSS_SK && $SK_NO_KEY; then
+ local numclients=${1:-$CLIENTCOUNT}
+ local clients=${CLIENTS:-$HOSTNAME}
+
+ # security ctx config for keyring
+ SK_NO_KEY=false
+ mkdir -p $SK_OM_PATH
+ mount -o bind $SK_OM_PATH /etc/request-key.d/
+ local lgssc_conf_line='create lgssc * * '
+ lgssc_conf_line+=$(which lgss_keyring)
+ lgssc_conf_line+=' %o %k %t %d %c %u %g %T %P %S'
+
+ local lgssc_conf_file="/etc/request-key.d/lgssc.conf"
+ echo "$lgssc_conf_line" > $lgssc_conf_file
+ [ -e $lgssc_conf_file ] ||
+ error_exit "Could not find key options in $lgssc_conf_file"
+
+ if ! local_mode; then
+ do_nodes $(comma_list $(all_nodes)) "mkdir -p \
+ $SK_OM_PATH"
+ do_nodes $(comma_list $(all_nodes)) "mount \
+ -o bind $SK_OM_PATH \
+ /etc/request-key.d/"
+ do_nodes $(comma_list $(all_nodes)) "rsync -aqv \
+ $HOSTNAME:$lgssc_conf_file \
+ $lgssc_conf_file >/dev/null 2>&1"
+ fi
+
+ # create shared key on all nodes
+ mkdir -p $SK_PATH/nodemap
+ rm -f $SK_PATH/$FSNAME.key $SK_PATH/nodemap/c*.key \
+ $SK_PATH/$FSNAME-*.key
+ # for nodemap testing each client may need own key,
+ # and S2S now requires keys as well, both for "client"
+ # and for "server"
+ if $SK_S2S; then
+ lgss_sk -t server -f$FSNAME -n $SK_S2SNMCLI \
+ -w $SK_PATH/$FSNAME-nmclient.key \
+ -d /dev/urandom >/dev/null 2>&1
+ lgss_sk -t mgs,server -f$FSNAME -n $SK_S2SNM \
+ -w $SK_PATH/$FSNAME-s2s-server.key \
+ -d /dev/urandom >/dev/null 2>&1
+ fi
+ # basic key create
+ lgss_sk -t server -f$FSNAME -w $SK_PATH/$FSNAME.key \
+ -d /dev/urandom >/dev/null 2>&1
+ # per-nodemap keys
+ for i in $(seq 0 $((numclients - 1))); do
+ lgss_sk -t server -f$FSNAME -n c$i \
+ -w $SK_PATH/nodemap/c$i.key -d /dev/urandom \
+ >/dev/null 2>&1
+ done
+ # Distribute keys
+ if ! local_mode; then
+ do_nodes $(comma_list $(all_nodes)) "rsync -av \
+ $HOSTNAME:$SK_PATH/ $SK_PATH >/dev/null 2>&1"
+ fi
+ # Set client keys to client type to generate prime P
+ if local_mode; then
+ do_nodes $(all_nodes) "lgss_sk -t client,server -m \
+ $SK_PATH/$FSNAME.key >/dev/null 2>&1"
+ else
+ do_nodes $clients "lgss_sk -t client -m \
+ $SK_PATH/$FSNAME.key >/dev/null 2>&1"
+ do_nodes $clients "find $SK_PATH/nodemap -name \*.key | \
+ xargs -IX lgss_sk -t client -m X >/dev/null 2>&1"
+ fi
+ # This is required for servers as well, if S2S in use
+ if $SK_S2S; then
+ do_nodes $(comma_list $(mdts_nodes)) \
+ "cp $SK_PATH/$FSNAME-s2s-server.key \
+ $SK_PATH/$FSNAME-s2s-client.key; lgss_sk \
+ -t client -m $SK_PATH/$FSNAME-s2s-client.key \
+ >/dev/null 2>&1"
+ do_nodes $(comma_list $(osts_nodes)) \
+ "cp $SK_PATH/$FSNAME-s2s-server.key \
+ $SK_PATH/$FSNAME-s2s-client.key; lgss_sk \
+ -t client -m $SK_PATH/$FSNAME-s2s-client.key \
+ >/dev/null 2>&1"
+ do_nodes $clients "lgss_sk -t client \
+ -m $SK_PATH/$FSNAME-nmclient.key \
+ >/dev/null 2>&1"
+ fi
+ # mount options for servers and clients
+ MGS_MOUNT_OPTS=$(add_sk_mntflag $MGS_MOUNT_OPTS)
+ MDS_MOUNT_OPTS=$(add_sk_mntflag $MDS_MOUNT_OPTS)
+ OST_MOUNT_OPTS=$(add_sk_mntflag $OST_MOUNT_OPTS)
+ MOUNT_OPTS=$(add_sk_mntflag $MOUNT_OPTS)
+ SEC=$SK_FLAVOR
+ fi
- if [ -n "$LGSS_KEYRING_DEBUG" ]; then
+ if [ -n "$LGSS_KEYRING_DEBUG" ]; then
lctl set_param -n \
- sptlrpc.gss.lgss_keyring.debug_level=$LGSS_KEYRING_DEBUG
- fi
- fi
+ sptlrpc.gss.lgss_keyring.debug_level=$LGSS_KEYRING_DEBUG
+ fi
}
cleanup_gss() {
- if $GSS; then
- stop_gss_daemons
- # maybe cleanup credential cache?
- fi
+ if $GSS; then
+ stop_gss_daemons
+ # maybe cleanup credential cache?
+ fi
+}
+
+cleanup_sk() {
+ if $GSS_SK; then
+ if $SK_S2S; then
+ do_node $(mgs_node) "$LCTL nodemap_del $SK_S2SNM"
+ do_node $(mgs_node) "$LCTL nodemap_del $SK_S2SNMCLI"
+ $RPC_MODE || echo "Sleeping for 10 sec for Nodemap.."
+ sleep 10
+ fi
+ stop_gss_daemons
+ $RPC_MODE || echo "Cleaning up Shared Key.."
+ do_nodes $(comma_list $(all_nodes)) "rm -f \
+ $SK_PATH/$FSNAME*.key $SK_PATH/nodemap/$FSNAME*.key"
+ # Remove the mount and clean up the files we added to SK_PATH
+ do_nodes $(comma_list $(all_nodes)) "umount \
+ /etc/request-key.d/"
+ do_nodes $(comma_list $(all_nodes)) "rm -f \
+ $SK_OM_PATH/lgssc.conf"
+ do_nodes $(comma_list $(all_nodes)) "rmdir $SK_OM_PATH"
+ SK_NO_KEY=true
+ fi
}
facet_svc() {
do_node $client "! grep -q $mnt' ' /proc/mounts ||
umount $mnt"
fi
- do_node $client $MOUNT_CMD $flags $opts $device $mnt || return 1
+ if $GSS_SK && ($SK_UNIQUE_NM || $SK_S2S); then
+ # Mount using nodemap key
+ local mountkey=$SK_PATH/$FSNAME-nmclient.key
+ if $SK_UNIQUE_NM; then
+ mountkey=$SK_PATH/nodemap/c0.key
+ fi
+ local prunedopts=$(echo $opts |
+ sed -e "s#skpath=[^,^ ]*#skpath=$mountkey#g")
+ do_node $client $MOUNT_CMD $flags $prunedopts $device $mnt ||
+ return 1
+ else
+ do_node $client $MOUNT_CMD $flags $opts $device $mnt ||
+ return 1
+ fi
set_default_debug_nodes $client
fi
}
+# Mount the file system on the MGS
+mount_mgs_client() {
+ do_facet mgs "mkdir -p $MOUNT"
+ zconf_mount $mgs_HOST $MOUNT $MOUNT_OPTS ||
+ error "unable to mount $MOUNT on MGS"
+}
+
+# Unmount the file system on the MGS
+umount_mgs_client() {
+ zconf_umount $mgs_HOST $MOUNT
+ do_facet mgs "rm -rf $MOUNT"
+}
+
# nodes is comma list
sanity_mount_check_nodes () {
local nodes=$1
local opts=${3:-$MOUNT_OPTS}
opts=${opts:+-o $opts}
local flags=${4:-$MOUNT_FLAGS}
-
local device=$MGSNID:/$FSNAME$FILESET
if [ -z "$mnt" -o -z "$FSNAME" ]; then
echo "Bad conf mount command: opt=$flags $opts dev=$device " \
echo "Starting client $clients: $flags $opts $device $mnt"
if [ -n "$FILESET" -a ! -n "$SKIP_FILESET" ]; then
- do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
- umount $mnt"
- do_nodes $clients $MOUNT_CMD $flags $opts $MGSNID:/$FSNAME \
- $mnt || return 1
+ if $GSS_SK && ($SK_UNIQUE_NM || $SK_S2S); then
+ # Mount with own nodemap key
+ local i=0
+ # Mount all server nodes first with per-NM keys
+ for nmclient in ${clients//,/ }; do
+# do_nodes $(comma_list $(all_server_nodes)) "lgss_sk -t server -l $SK_PATH/nodemap/c$i.key -n c$i"
+ do_nodes $(comma_list $(all_server_nodes)) "lgss_sk -t server -l $SK_PATH/nodemap/c$i.key"
+ i=$((i + 1))
+ done
+ # set perms for per-nodemap keys else permission denied
+ do_nodes $(comma_list $(all_nodes)) \
+ "keyctl show | grep lustre | cut -c1-11 |
+ sed -e 's/ //g;' |
+ xargs -IX keyctl setperm X 0x3f3f3f3f"
+ local mountkey=$SK_PATH/$FSNAME-nmclient.key
+ i=0
+ for nmclient in ${clients//,/ }; do
+ if $SK_UNIQUE_NM; then
+ mountkey=$SK_PATH/nodemap/c$i.key
+ fi
+ do_node $nmclient "! grep -q $mnt' ' \
+ /proc/mounts || umount $mnt"
+ local prunedopts=$(add_sk_mntflag $prunedopts);
+ prunedopts=$(echo $prunedopts | sed -e \
+ "s#skpath=[^ ^,]*#skpath=$mountkey#g")
+ set -x
+ do_nodes $(comma_list $(all_server_nodes)) \
+ "keyctl show"
+ set +x
+ do_node $nmclient $MOUNT_CMD $flags \
+ $prunedopts $MGSNID:/$FSNAME $mnt ||
+ return 1
+ i=$((i + 1))
+ done
+ else
+ do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
+ umount $mnt"
+ do_nodes $clients $MOUNT_CMD $flags $opts \
+ $MGSNID:/$FSNAME $mnt || return 1
+ fi
#disable FILESET if not supported
do_nodes $clients lctl get_param -n \
mdc.$FSNAME-MDT0000*.import | grep -q subtree ||
umount $mnt"
fi
- do_nodes $clients "
+ if $GSS_SK && ($SK_UNIQUE_NM || $SK_S2S); then
+ # Mount with nodemap key
+ local i=0
+ local mountkey=$SK_PATH/$FSNAME-nmclient.key
+ for nmclient in ${clients//,/ }; do
+ if $SK_UNIQUE_NM; then
+ mountkey=$SK_PATH/nodemap/c$i.key
+ fi
+ local prunedopts=$(echo $opts | sed -e \
+ "s#skpath=[^ ^,]*#skpath=$mountkey#g");
+ do_node $nmclient "! grep -q $mnt' ' /proc/mounts ||
+ umount $mnt"
+ do_node $nmclient "
+ running=\\\$(mount | grep -c $mnt' ');
+ rc=0;
+ if [ \\\$running -eq 0 ] ; then
+ mkdir -p $mnt;
+ $MOUNT_CMD $flags $prunedopts $device $mnt;
+ rc=\\\$?;
+ else
+ lustre_mnt_count=\\\$(mount | grep $mnt' ' | \
+ grep 'type lustre' | wc -l);
+ if [ \\\$running -ne \\\$lustre_mnt_count ] ; then
+ echo zconf_mount_clients FAILED: \
+ mount count \\\$running, not matching \
+ with mount count of 'type lustre' \
+ \\\$lustre_mnt_count;
+ rc=1;
+ fi;
+ fi;
+ exit \\\$rc" || return ${PIPESTATUS[0]}
+
+ i=$((i + 1))
+ done
+ else
+
+ local tmpopts=$opts
+ if $SHARED_KEY; then
+ tmpopts=$(add_sk_mntflag $opts)
+ fi
+ do_nodes $clients "
running=\\\$(mount | grep -c $mnt' ');
rc=0;
if [ \\\$running -eq 0 ] ; then
- mkdir -p $mnt;
- $MOUNT_CMD $flags $opts $device $mnt;
- rc=\\\$?;
+ mkdir -p $mnt;
+ $MOUNT_CMD $flags $tmpopts $device $mnt;
+ rc=\\\$?;
fi;
exit \\\$rc" || return ${PIPESTATUS[0]}
+ fi
echo "Started clients $clients: "
do_nodes $clients "mount | grep $mnt' '"
sync_all_data() {
do_nodes $(comma_list $(mdts_nodes)) \
- "lctl set_param -n osd*.*MDT*.force_sync=1"
+ "lctl set_param -n os[cd]*.*MDT*.force_sync=1"
do_nodes $(comma_list $(osts_nodes)) \
"lctl set_param -n osd*.*OS*.force_sync=1" 2>&1 |
grep -v 'Found no match'
mds2sync=$(comma_list $mds2sync)
# sync MDS transactions
- do_nodes $mds2sync "$LCTL set_param -n osd*.*MD*.force_sync 1"
+ do_nodes $mds2sync "$LCTL set_param -n os[cd]*.*MD*.force_sync 1"
# wait till all changes are sent and commmitted by OSTs
# for ldiskfs space is released upon execution, but DMU
wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
error "wait_update OSTs up on client failed"
- cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
- awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
+ cmd="$LCTL get_param osp.$FSNAME-OST*-MDT0000.prealloc_last_id |
+ awk '/=[1-9][0-9]/ { c += 1 } END { printf \\\"%d\\\", c }'"
wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
- error "wait_update OSTs up on MDT failed"
+ error "wait_update OSTs up on MDT0000 failed"
}
wait_destroy_complete () {
fi
}
-
clients_up() {
# not every config has many clients
sleep 1
clients_up || error "post-failover stat: $?"
}
-do_lmc() {
- echo There is no lmc. This is mountconf, baby.
- exit 1
-}
-
host_nids_address() {
- local nodes=$1
- local kind=$2
+ local nodes=$1
+ local net=${2:-"."}
- if [ -n "$kind" ]; then
- nids=$(do_nodes $nodes "$LCTL list_nids | grep $kind | cut -f 1 -d '@'")
- else
- nids=$(do_nodes $nodes "$LCTL list_nids all | cut -f 1 -d '@'")
- fi
- echo $nids
+ do_nodes $nodes "$LCTL list_nids | grep $net | cut -f 1 -d @"
}
h2name_or_ip() {
fi
}
-h2ptl() {
- if [ "$1" = "'*'" ]; then echo \'*\'; else
- ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | \
- awk '{print $1}'`
- if [ -z "$ID" ]; then
- echo "Could not get a ptl id for $1..."
- exit 1
- fi
- echo $ID"@ptl"
+h2nettype() {
+ if [[ -n "$NETTYPE" ]]; then
+ h2name_or_ip "$1" "$NETTYPE"
+ else
+ h2name_or_ip "$1" "$2"
fi
}
-declare -fx h2ptl
+declare -fx h2nettype
+# Wrapper function to print the deprecation warning
h2tcp() {
- h2name_or_ip "$1" "tcp"
-}
-declare -fx h2tcp
-
-h2elan() {
- if [ "$1" = "'*'" ]; then echo \'*\'; else
- if type __h2elan >/dev/null 2>&1; then
- ID=$(__h2elan $1)
- else
- ID=`echo $1 | sed 's/[^0-9]*//g'`
- fi
- echo $ID"@elan"
+ echo "h2tcp: deprecated, use h2nettype instead" 1>&2
+ if [[ -n "$NETTYPE" ]]; then
+ h2nettype "$@"
+ else
+ h2nettype "$1" "tcp"
fi
}
-declare -fx h2elan
+# Wrapper function to print the deprecation warning
h2o2ib() {
- h2name_or_ip "$1" "o2ib"
+ echo "h2o2ib: deprecated, use h2nettype instead" 1>&2
+ if [[ -n "$NETTYPE" ]]; then
+ h2nettype "$@"
+ else
+ h2nettype "$1" "o2ib"
+ fi
}
-declare -fx h2o2ib
# This enables variables in cfg/"setup".sh files to support the pdsh HOSTLIST
# expressions format. As a bonus we can then just pass in those variables
}
cleanupall() {
- nfs_client_mode && return
+ nfs_client_mode && return
cifs_client_mode && return
- stopall $*
- cleanup_echo_devs
+ stopall $*
+ cleanup_echo_devs
- unload_modules
- cleanup_gss
+ unload_modules
+ cleanup_sk
+ cleanup_gss
}
combined_mgs_mds () {
var=${facet}failover_HOST
if [ -n "${!var}" ] && [ ${!var} != $(facet_host $facet) ]; then
- opts+=" --failnode=$(h2$NETTYPE ${!var})"
+ opts+=" --failnode=$(h2nettype ${!var})"
fi
opts+=${TIMEOUT:+" --param=sys.timeout=$TIMEOUT"}
done
}
+__touch_device()
+{
+ local facet_type=$1 # mgs || mds || ost
+ local facet_num=$2
+ local facet=${1}${2}
+ local device
+
+ case "$(facet_fstype $facet)" in
+ ldiskfs)
+ device=$(${facet_type}devname $facet_num)
+ ;;
+ zfs)
+ device=$(${facet_type}vdevname $facet_num)
+ ;;
+ *)
+ error "Unhandled filesystem type"
+ ;;
+ esac
+
+ do_facet $facet "[ -e \"$device\" ]" && return
+
+ # Note: the following check only works with absolute paths
+ [[ ! "$device" =~ ^/dev/ ]] || [[ "$device" =~ ^/dev/shm/ ]] ||
+ error "$facet: device '$device' does not exist"
+
+ do_facet $facet "touch \"${device}\""
+}
+
format_mgs() {
local quiet
fi
echo "Format mgs: $(mgsdevname)"
reformat_external_journal mgs
+
+ # touch "device" in case it is a loopback file for testing and needs to
+ # be created. mkfs.lustre doesn't do this to avoid accidentally writing
+ # to non-existent files in /dev if the admin made a typo during setup
+ __touch_device mgs
+
add mgs $(mkfs_opts mgs $(mgsdevname)) $(mountfs_opts mgs) --reformat \
$(mgsdevname) $(mgsvdevname) ${quiet:+>/dev/null} || exit 10
}
fi
echo "Format mds$num: $(mdsdevname $num)"
reformat_external_journal mds$num
+
+ __touch_device mds $num
+
add mds$num $(mkfs_opts mds$num $(mdsdevname ${num})) \
$(mountfs_opts mds$num) --reformat $(mdsdevname $num) \
$(mdsvdevname $num) ${quiet:+>/dev/null} || exit 10
fi
echo "Format ost$num: $(ostdevname $num)"
reformat_external_journal ost$num
+
+ __touch_device ost $num
+
add ost$num $(mkfs_opts ost$num $(ostdevname ${num})) \
$(mountfs_opts ost$num) --reformat $(ostdevname $num) \
$(ostvdevname ${num}) ${quiet:+>/dev/null} || exit 10
}
formatall() {
- stopall
+ stopall -f
# Set hostid for ZFS/SPL zpool import protection
# (Assumes MDS version is also OSS version)
if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.8.54) ];
}
mount_client() {
- grep " $1 " /proc/mounts || zconf_mount $HOSTNAME $*
+ grep " $1 " /proc/mounts || zconf_mount $HOSTNAME $*
}
umount_client() {
- grep " $1 " /proc/mounts && zconf_umount `hostname` $*
+ grep " $1 " /proc/mounts && zconf_umount $HOSTNAME $*
}
# return value:
remount_client()
{
- zconf_umount `hostname` $1 || error "umount failed"
- zconf_mount `hostname` $1 || error "mount failed"
+ zconf_umount $HOSTNAME $1 || error "umount failed"
+ zconf_mount $HOSTNAME $1 || error "mount failed"
}
writeconf_facet() {
return $rc
}
+mountmgs() {
+ if ! combined_mgs_mds ; then
+ start mgs $(mgsdevname) $MGS_MOUNT_OPTS
+ fi
+}
+
+mountmds() {
+ for num in $(seq $MDSCOUNT); do
+ DEVNAME=$(mdsdevname $num)
+ start mds$num $DEVNAME $MDS_MOUNT_OPTS
+
+ # We started mds, now we should set failover variables properly.
+ # Set mds${num}failover_HOST if unset (the default
+ # failnode).
+ local varname=mds${num}failover_HOST
+ if [ -z "${!varname}" ]; then
+ eval mds${num}failover_HOST=$(facet_host mds$num)
+ fi
+
+ if [ $IDENTITY_UPCALL != "default" ]; then
+ switch_identity $num $IDENTITY_UPCALL
+ fi
+ done
+}
+
+mountoss() {
+ for num in $(seq $OSTCOUNT); do
+ DEVNAME=$(ostdevname $num)
+ start ost$num $DEVNAME $OST_MOUNT_OPTS
+
+ # We started ost$num, now we should set ost${num}failover
+ # variable properly. Set ost${num}failover_HOST if it is not
+ # set (the default failnode).
+ varname=ost${num}failover_HOST
+ if [ -z "${!varname}" ]; then
+ eval ost${num}failover_HOST=$(facet_host ost${num})
+ fi
+
+ done
+}
+
+mountcli() {
+ [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
+ if [ ! -z $arg1 ]; then
+ [ "$arg1" = "server_only" ] && return
+ fi
+ mount_client $MOUNT
+ if [ -n "$CLIENTS" ]; then
+ zconf_mount_clients $CLIENTS $MOUNT
+ fi
+ clients_up
+
+ if [ "$MOUNT_2" ]; then
+ mount_client $MOUNT2
+ if [ -n "$CLIENTS" ]; then
+ zconf_mount_clients $CLIENTS $MOUNT2
+ fi
+ fi
+}
+
+sk_nodemap_setup() {
+ local sk_map_name=${1:-$SK_S2SNM}
+ local sk_map_nodes=${2:-$HOSTNAME}
+ do_node $(mgs_node) "$LCTL nodemap_add $sk_map_name"
+ for servernode in $sk_map_nodes; do
+ local nids=$(do_nodes $servernode "$LCTL list_nids")
+ for nid in $nids; do
+ do_node $(mgs_node) "$LCTL nodemap_add_range --name \
+ $sk_map_name --range $nid"
+ done
+ done
+}
+
setupall() {
local arg1=$1
load_modules
+ init_gss
+
if [ -z "$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
-
- for num in `seq $MDSCOUNT`; do
- DEVNAME=$(mdsdevname $num)
- start mds$num $DEVNAME $MDS_MOUNT_OPTS
-
- # We started mds, now we should set failover variables properly.
- # Set mds${num}failover_HOST if it is not set (the default failnode).
- local varname=mds${num}failover_HOST
- if [ -z "${!varname}" ]; then
- eval mds${num}failover_HOST=$(facet_host mds$num)
- fi
- if [ $IDENTITY_UPCALL != "default" ]; then
- switch_identity $num $IDENTITY_UPCALL
- fi
- done
- for num in `seq $OSTCOUNT`; do
- DEVNAME=$(ostdevname $num)
- start ost$num $DEVNAME $OST_MOUNT_OPTS
-
- # We started ost$num, now we should set ost${num}failover variable properly.
- # Set ost${num}failover_HOST if it is not set (the default failnode).
- varname=ost${num}failover_HOST
- if [ -z "${!varname}" ]; then
- eval ost${num}failover_HOST=$(facet_host ost${num})
- fi
-
- done
- fi
-
- init_gss
-
- # wait a while to allow sptlrpc configuration be propogated to targets,
- # only needed when mounting new target devices.
- if $GSS; then
- sleep 10
- fi
-
- [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
-
- if [ ! -z $arg1 ]; then
- [ "$arg1" = "server_only" ] && return
+ if $SK_MOUNTED; then
+ echo "Shared Key file system already mounted"
+ else
+ mountmgs
+ mountmds
+ mountoss
+ if $SHARED_KEY; then
+ export SK_MOUNTED=true
+ fi
+ fi
+ if $GSS_SK; then
+ echo "GSS_SK: setting kernel keyring perms"
+ do_nodes $(comma_list $(all_nodes)) \
+ "keyctl show | grep lustre | cut -c1-11 |
+ sed -e 's/ //g;' |
+ xargs -IX keyctl setperm X 0x3f3f3f3f"
+
+ if $SK_S2S; then
+ # Need to start one nodemap for servers,
+ # and one for clients.
+ sk_nodemap_setup $SK_S2SNM \
+ $(comma_list $(all_server_nodes))
+ mountcli
+ sk_nodemap_setup $SK_S2SNMCLI \
+ ${CLIENTS:-$HOSTNAME}
+ echo "Nodemap set up for SK S2S, remounting."
+ stopall
+ mountmgs
+ mountmds
+ mountoss
+ fi
+ fi
fi
- mount_client $MOUNT
- [ -n "$CLIENTS" ] && zconf_mount_clients $CLIENTS $MOUNT
- clients_up
-
- if [ "$MOUNT_2" ]; then
- mount_client $MOUNT2
- [ -n "$CLIENTS" ] && zconf_mount_clients $CLIENTS $MOUNT2
- fi
+ # wait a while to allow sptlrpc configuration be propogated to targets,
+ # only needed when mounting new target devices.
+ if $GSS; then
+ sleep 10
+ fi
- init_param_vars
+ mountcli
+ init_param_vars
- # by remounting mdt before ost, initial connect from mdt to ost might
- # timeout because ost is not ready yet. wait some time to its fully
- # recovery. initial obd_connect timeout is 5s; in GSS case it's preceeded
- # by a context negotiation rpc with $TIMEOUT.
- # FIXME better by monitoring import status.
- if $GSS; then
- set_flavor_all $SEC
- sleep $((TIMEOUT + 5))
- else
- sleep 5
- fi
+ # by remounting mdt before ost, initial connect from mdt to ost might
+ # timeout because ost is not ready yet. wait some time to its fully
+ # recovery. initial obd_connect timeout is 5s; in GSS case it's
+ # preceeded by a context negotiation rpc with $TIMEOUT.
+ # FIXME better by monitoring import status.
+ if $GSS; then
+ if $GSS_SK; then
+ set_rule $FSNAME any cli2mdt $SK_FLAVOR
+ set_rule $FSNAME any cli2ost $SK_FLAVOR
+ wait_flavor cli2mdt $SK_FLAVOR
+ wait_flavor cli2ost $SK_FLAVOR
+ else
+ set_flavor_all $SEC
+ fi
+ sleep $((TIMEOUT + 5))
+ else
+ sleep 5
+ fi
}
mounted_lustre_filesystems() {
return 0
fi
- local myMGS_host=$mgs_HOST
- if [ "$NETTYPE" = "ptl" ]; then
- myMGS_host=$(h2ptl $mgs_HOST | sed -e s/@ptl//)
- fi
-
echo Checking config lustre mounted on $mntpt
local mgshost=$(mount | grep " $mntpt " | awk -F@ '{print $1}')
mgshost=$(echo $mgshost | awk -F: '{print $1}')
-# if [ "$mgshost" != "$myMGS_host" ]; then
-# log "Bad config file: lustre is mounted with mgs $mgshost, but mgs_HOST=$mgs_HOST, NETTYPE=$NETTYPE
-# Please use correct config or set mds_HOST correctly!"
-# fi
-
}
check_config_clients () {
fi
init_gss
- if $GSS; then
+ if $GSS_SK; then
+ set_flavor_all null
+ elif $GSS; then
set_flavor_all $SEC
fi
}
check_network() {
- local host=$1
- local max=$2
- local sleep=${3:-5}
+ local host=$1
+ local max=$2
+ local sleep=${3:-5}
- echo `date +"%H:%M:%S (%s)"` waiting for $host network $max secs ...
- if ! wait_for_function --quiet "ping -c 1 -w 3 $host" $max $sleep ; then
- echo "Network not available!"
- exit 1
- fi
+ [ "$host" = "$HOSTNAME" ] && return 0
+
+ echo "$(date +'%H:%M:%S (%s)') waiting for $host network $max secs ..."
+ if ! wait_for_function --quiet "ping -c 1 -w 3 $host" $max $sleep ; then
+ echo "Network not available!"
+ exit 1
+ fi
- echo `date +"%H:%M:%S (%s)"` network interface is UP
+ echo "$(date +'%H:%M:%S (%s)') network interface is UP"
}
no_dsh() {
# Test interface
##################################
+# usage: stack_trap arg sigspec
+#
+# stack_trap() behaves like bash's built-in trap, except that it "stacks" the
+# command ``arg`` on top of previously defined commands for ``sigspec`` instead
+# of overwriting them.
+# stacked traps are executed in reverse order of their registration
+#
+# arg and sigspec have the same meaning as in man (1) trap
+stack_trap()
+{
+ local arg="$1"
+ local sigspec="$2"
+
+ local cmd="$(trap -p $sigspec)"
+
+ cmd="${cmd#trap -- \'}"
+ cmd="${cmd%\'*}"
+ [ -n "$cmd" ] && cmd="; $cmd"
+ cmd="${arg}$cmd"
+
+ trap "$cmd" $sigspec
+}
+
error_noexit() {
report_error "$@"
}
echo -n $nodes_sort
}
+# Get name of the active MGS node.
+mgs_node () {
+ echo -n $(facets_nodes $(get_facets MGS))
+}
+
# Get all of the active MDS nodes.
mdts_nodes () {
echo -n $(facets_nodes $(get_facets MDS))
all_mdts_nodes () {
local host
local failover_host
- local nodes
+ local nodes="${mds_HOST} ${mdsfailover_HOST}"
local nodes_sort
local i
all_osts_nodes () {
local host
local failover_host
- local nodes
+ local nodes="${ost_HOST} ${ostfailover_HOST}"
local nodes_sort
local i
setstripe_nfsserver () {
local dir=$1
+ local nfsexportdir=$2
+ shift
+ shift
- local nfsserver=$(awk '"'$dir'" ~ $2 && $3 ~ "nfs" && $2 != "/" \
- { print $1 }' /proc/mounts | cut -f 1 -d : | head -n1)
+ local -a nfsexport=($(awk '"'$dir'" ~ $2 && $3 ~ "nfs" && $2 != "/" \
+ { print $1 }' /proc/mounts | cut -f 1 -d :))
- [ -z $nfsserver ] && echo "$dir is not nfs mounted" && return 1
+ # check that only one nfs mounted
+ [[ -z $nfsexport ]] && echo "$dir is not nfs mounted" && return 1
+ (( ${#nfsexport[@]} == 1 )) ||
+ error "several nfs mounts found for $dir: ${nfsexport[@]} !"
- do_nodev $nfsserver lfs setstripe "$@"
+ do_nodev ${nfsexport[0]} lfs setstripe $nfsexportdir "$@"
}
# Check and add a test group.
save_lustre_params() {
local facets=$1
local facet
- local nodes
- local node
+ local facet_svc
for facet in ${facets//,/ }; do
- node=$(facet_active_host $facet)
- [[ *\ $node\ * = " $nodes " ]] && continue
- nodes="$nodes $node"
-
- do_node $node "$LCTL get_param $2 |
- while read s; do echo $facet \\\$s; done"
+ facet_svc=$(facet_svc $facet)
+ do_facet $facet \
+ "params=\\\$($LCTL get_param $2);
+ [[ -z \\\"$facet_svc\\\" ]] && param= ||
+ param=\\\$(grep $facet_svc <<< \\\"\\\$params\\\");
+ [[ -z \\\$param ]] && param=\\\"\\\$params\\\";
+ while read s; do echo $facet \\\$s;
+ done <<< \\\"\\\$param\\\""
done
}
}
get_clientosc_proc_path() {
- echo "${1}-osc-*"
+ echo "${1}-osc-ffff*"
}
# If the 2.0 MDS was mounted on 1.8 device, then the OSC and LOV names
local mdt_label=$(convert_facet2label $mds_facet)
local mdt_index=$(echo $mdt_label | sed -e 's/^.*-//')
- if [ $(lustre_version_code $mds_facet) -le $(version_code 1.8.0) ] ||
- mds_on_old_device $mds_facet; then
- echo "${ost_label}-osc"
- elif [[ $ost_label = *OST* ]]; then
+ if [[ $ost_label = *OST* ]]; then
echo "${ost_label}-osc-${mdt_index}"
else
echo "${ost_label}-osp-${mdt_index}"
do_nodesv $list \
"$LCTL dk > ${prefix}.debug_log.\\\$(hostname -s).${suffix};
dmesg > ${prefix}.dmesg.\\\$(hostname -s).${suffix}"
+
if [ ! -f $LOGDIR/shared ]; then
do_nodes $list rsync -az "${prefix}.*.${suffix}" $HOSTNAME:$LOGDIR
fi
}
get_clients_mount_count () {
- local clients=${CLIENTS:-`hostname`}
+ local clients=${CLIENTS:-$HOSTNAME}
- # we need to take into account the clients mounts and
- # exclude mds/ost mounts if any;
- do_nodes $clients cat /proc/mounts | grep lustre | grep $MOUNT | wc -l
+ # we need to take into account the clients mounts and
+ # exclude mds/ost mounts if any;
+ do_nodes $clients cat /proc/mounts | grep lustre |
+ grep -w $MOUNT | wc -l
}
# gss functions
PROC_CLI="srpc_info"
+PROC_CON="srpc_contexts"
combination()
{
}
calc_connection_cnt() {
- local dir=$1
+ local dir=$1
- # MDT->MDT = 2 * C(M, 2)
- # MDT->OST = M * O
- # CLI->OST = C * O
- # CLI->MDT = C * M
- comb_m2=$(combination $MDSCOUNT 2)
+ # MDT->MDT = 2 * C(M, 2)
+ # MDT->OST = M * O
+ # CLI->OST = C * O
+ # CLI->MDT = C * M
+ comb_m2=$(combination $MDSCOUNT 2)
- local num_clients=$(get_clients_mount_count)
+ local num_clients=$(get_clients_mount_count)
- local cnt_mdt2mdt=$((comb_m2 * 2))
- local cnt_mdt2ost=$((MDSCOUNT * OSTCOUNT))
- local cnt_cli2ost=$((num_clients * OSTCOUNT))
- local cnt_cli2mdt=$((num_clients * MDSCOUNT))
- local cnt_all2ost=$((cnt_mdt2ost + cnt_cli2ost))
- local cnt_all2mdt=$((cnt_mdt2mdt + cnt_cli2mdt))
- local cnt_all2all=$((cnt_mdt2ost + cnt_mdt2mdt + cnt_cli2ost + cnt_cli2mdt))
+ local cnt_mdt2mdt=$((comb_m2 * 2))
+ local cnt_mdt2ost=$((MDSCOUNT * OSTCOUNT))
+ local cnt_cli2ost=$((num_clients * OSTCOUNT))
+ local cnt_cli2mdt=$((num_clients * MDSCOUNT))
+ if is_mounted $MOUNT2; then
+ cnt_cli2mdt=$((cnt_cli2mdt * 2))
+ cnt_cli2ost=$((cnt_cli2ost * 2))
+ fi
+ if local_mode; then
+ cnt_mdt2mdt=0
+ cnt_mdt2ost=0
+ cnt_cli2ost=2
+ cnt_cli2mdt=1
+ fi
+ local cnt_all2ost=$((cnt_mdt2ost + cnt_cli2ost))
+ local cnt_all2mdt=$((cnt_mdt2mdt + cnt_cli2mdt))
+ local cnt_all2all=$((cnt_mdt2ost + cnt_mdt2mdt \
+ + cnt_cli2ost + cnt_cli2mdt))
- local var=cnt_$dir
- local res=${!var}
+ local var=cnt_$dir
+ local res=${!var}
- echo $res
+ echo $res
}
set_rule()
do_facet mgs "$LCTL conf_param $cmd"
}
+count_contexts()
+{
+ local output=$1
+ local total_ctx=$(echo "$output" | grep -c "expire.*key.*hdl")
+ echo $total_ctx
+}
+
count_flvr()
{
local output=$1
local flavor=$1
local cnt
- local clients=${CLIENTS:-`hostname`}
+ local clients=${CLIENTS:-$HOSTNAME}
for c in ${clients//,/ }; do
- output=`do_node $c lctl get_param -n mdc.*-MDT*-mdc-*.$PROC_CLI 2>/dev/null`
- tmpcnt=`count_flvr "$output" $flavor`
- cnt=$((cnt + tmpcnt))
+ local output=$(do_node $c lctl get_param -n \
+ mdc.*-*-mdc-*.$PROC_CLI 2>/dev/null)
+ local tmpcnt=$(count_flvr "$output" $flavor)
+ if $GSS_SK && [ $flavor != "null" ]; then
+ # tmpcnt=min(contexts,flavors) to ensure SK context is on
+ output=$(do_node $c lctl get_param -n \
+ mdc.*-MDT*-mdc-*.$PROC_CON 2>/dev/null)
+ local outcon=$(count_contexts "$output")
+ if [ "$outcon" -lt "$tmpcnt" ]; then
+ tmpcnt=$outcon
+ fi
+ fi
+ cnt=$((cnt + tmpcnt))
done
echo $cnt
}
local flavor=$1
local cnt
- local clients=${CLIENTS:-`hostname`}
+ local clients=${CLIENTS:-$HOSTNAME}
for c in ${clients//,/ }; do
- output=`do_node $c lctl get_param -n osc.*OST*-osc-[^M][^D][^T]*.$PROC_CLI 2>/dev/null`
- tmpcnt=`count_flvr "$output" $flavor`
+ local output=$(do_node $c lctl get_param -n \
+ osc.*OST*-osc-[^M][^D][^T]*.$PROC_CLI 2>/dev/null)
+ local tmpcnt=$(count_flvr "$output" $flavor)
+ if $GSS_SK && [ $flavor != "null" ]; then
+ # tmpcnt=min(contexts,flavors) to ensure SK context is on
+ output=$(do_node $c lctl get_param -n \
+ osc.*OST*-osc-[^M][^D][^T]*.$PROC_CON 2>/dev/null)
+ local outcon=$(count_contexts "$output")
+ if [ "$outcon" -lt "$tmpcnt" ]; then
+ tmpcnt=$outcon
+ fi
+ fi
cnt=$((cnt + tmpcnt))
done
echo $cnt
fi
for num in `seq $MDSCOUNT`; do
- output=`do_facet mds$num lctl get_param -n mdc.*-MDT*-mdc[0-9]*.$PROC_CLI 2>/dev/null`
- tmpcnt=`count_flvr "$output" $flavor`
+ local output=$(do_facet mds$num lctl get_param -n \
+ osp.*-MDT*osp-MDT*.$PROC_CLI 2>/dev/null)
+ local tmpcnt=$(count_flvr "$output" $flavor)
+ if $GSS_SK && [ $flavor != "null" ]; then
+ # tmpcnt=min(contexts,flavors) to ensure SK context is on
+ output=$(do_facet mds$num lctl get_param -n \
+ osp.*-MDT*osp-MDT*.$PROC_CON 2>/dev/null)
+ local outcon=$(count_contexts "$output")
+ if [ "$outcon" -lt "$tmpcnt" ]; then
+ tmpcnt=$outcon
+ fi
+ fi
cnt=$((cnt + tmpcnt))
done
echo $cnt;
for num in `seq $MDSCOUNT`; do
mdtosc=$(get_mdtosc_proc_path mds$num)
mdtosc=${mdtosc/-MDT*/-MDT\*}
- output=$(do_facet mds$num lctl get_param -n \
- osc.$mdtosc.$PROC_CLI 2>/dev/null)
- tmpcnt=`count_flvr "$output" $flavor`
+ local output=$(do_facet mds$num lctl get_param -n \
+ osc.$mdtosc.$PROC_CLI 2>/dev/null)
+ local tmpcnt=$(count_flvr "$output" $flavor)
+ if $GSS_SK && [ $flavor != "null" ]; then
+ # tmpcnt=min(contexts,flavors) to ensure SK context is on
+ output=$(do_facet mds$num lctl get_param -n \
+ osc.$mdtosc.$PROC_CON 2>/dev/null)
+ local outcon=$(count_contexts "$output")
+ if [ "$outcon" -lt "$tmpcnt" ]; then
+ tmpcnt=$outcon
+ fi
+ fi
cnt=$((cnt + tmpcnt))
done
echo $cnt;
{
local flavor=$1
- output=`do_facet client lctl get_param -n mgc.*.$PROC_CLI 2>/dev/null`
+ local output=$(do_facet client lctl get_param -n mgc.*.$PROC_CLI \
+ 2>/dev/null)
count_flvr "$output" $flavor
}
wait_flavor()
{
- local dir=$1 # from to
- local flavor=$2 # flavor expected
- local expect=${3:-$(calc_connection_cnt $dir)} # number expected
-
- local res=0
-
- for ((i=0;i<20;i++)); do
- echo -n "checking $dir..."
- res=$(do_check_flavor $dir $flavor)
- echo "found $res/$expect $flavor connections"
- [ $res -ge $expect ] && return 0
- sleep 4
- done
+ local dir=$1 # from to
+ local flavor=$2 # flavor expected
+ local expect=${3:-$(calc_connection_cnt $dir)} # number expected
+ local WAITFLAVOR_MAX=20 # how many retries before abort?
+
+ local res=0
+ for ((i = 0; i < $WAITFLAVOR_MAX; i++)); do
+ echo -n "checking $dir..."
+ res=$(do_check_flavor $dir $flavor)
+ echo "found $res/$expect $flavor connections"
+ [ $res -ge $expect ] && return 0
+ sleep 4
+ done
- echo "Error checking $flavor of $dir: expect $expect, actual $res"
- return 1
+ echo "Error checking $flavor of $dir: expect $expect, actual $res"
+# echo "Dumping additional logs for SK debug.."
+ do_nodes $(comma_list $(all_server_nodes)) "keyctl show"
+ if $dump; then
+ gather_logs $(comma_list $(nodes_list))
+ fi
+ return 1
}
restore_to_default_flavor()
{
- local proc="mgs.MGS.live.$FSNAME"
+ local proc="mgs.MGS.live.$FSNAME"
- echo "restoring to default flavor..."
+ echo "restoring to default flavor..."
- nrule=`do_facet mgs lctl get_param -n $proc 2>/dev/null | grep ".srpc.flavor." | wc -l`
+ local nrule=$(do_facet mgs lctl get_param -n $proc 2>/dev/null |
+ grep ".srpc.flavor" | wc -l)
- # remove all existing rules if any
- if [ $nrule -ne 0 ]; then
- echo "$nrule existing rules"
- for rule in `do_facet mgs lctl get_param -n $proc 2>/dev/null | grep ".srpc.flavor."`; do
- echo "remove rule: $rule"
- spec=`echo $rule | awk -F = '{print $1}'`
- do_facet mgs "$LCTL conf_param -d $spec"
- done
- fi
+ # remove all existing rules if any
+ if [ $nrule -ne 0 ]; then
+ echo "$nrule existing rules"
+ for rule in $(do_facet mgs lctl get_param -n $proc 2>/dev/null |
+ grep ".srpc.flavor."); do
+ echo "remove rule: $rule"
+ spec=`echo $rule | awk -F = '{print $1}'`
+ do_facet mgs "$LCTL conf_param -d $spec"
+ done
+ fi
- # verify no rules left
- nrule=`do_facet mgs lctl get_param -n $proc 2>/dev/null | grep ".srpc.flavor." | wc -l`
- [ $nrule -ne 0 ] && error "still $nrule rules left"
+ # verify no rules left
+ nrule=$(do_facet mgs lctl get_param -n $proc 2>/dev/null |
+ grep ".srpc.flavor." | wc -l)
+ [ $nrule -ne 0 ] && error "still $nrule rules left"
- # wait for default flavor to be applied
- # currently default flavor for all connections are 'null'
- wait_flavor all2all null
- echo "now at default flavor settings"
+ # wait for default flavor to be applied
+ if $GSS_SK; then
+ if $SK_S2S; then
+ set_rule $FSNAME any any $SK_FLAVOR
+ wait_flavor all2all $SK_FLAVOR
+ else
+ set_rule $FSNAME any cli2mdt $SK_FLAVOR
+ set_rule $FSNAME any cli2ost $SK_FLAVOR
+ wait_flavor cli2mdt $SK_FLAVOR
+ wait_flavor cli2ost $SK_FLAVOR
+ fi
+ echo "GSS_SK now at default flavor: $SK_FLAVOR"
+ else
+ wait_flavor all2all null
+ fi
}
set_flavor_all()
{
- local flavor=${1:-null}
+ local flavor=${1:-null}
- echo "setting all flavor to $flavor"
+ echo "setting all flavor to $flavor"
- # FIXME need parameter to this fn
- # and remove global vars
- local cnt_all2all=$(calc_connection_cnt all2all)
+ # FIXME need parameter to this fn
+ # and remove global vars
+ local cnt_all2all=$(calc_connection_cnt all2all)
- local res=$(do_check_flavor all2all $flavor)
- if [ $res -eq $cnt_all2all ]; then
- echo "already have total $res $flavor connections"
- return
- fi
+ local res=$(do_check_flavor all2all $flavor)
+ if [ $res -eq $cnt_all2all ]; then
+ echo "already have total $res $flavor connections"
+ return
+ fi
- echo "found $res $flavor out of total $cnt_all2all connections"
- restore_to_default_flavor
+ echo "found $res $flavor out of total $cnt_all2all connections"
+ restore_to_default_flavor
- [[ $flavor = null ]] && return 0
+ [[ $flavor = null ]] && return 0
- set_rule $FSNAME any any $flavor
- wait_flavor all2all $flavor
+ if $GSS_SK && [ $flavor != "null" ]; then
+ if $SK_S2S; then
+ set_rule $FSNAME any any $flavor
+ wait_flavor all2all $flavor
+ else
+ set_rule $FSNAME any cli2mdt $flavor
+ set_rule $FSNAME any cli2ost $flavor
+ set_rule $FSNAME any mdt2ost null
+ set_rule $FSNAME any mdt2mdt null
+ wait_flavor cli2mdt $flavor
+ wait_flavor cli2ost $flavor
+ fi
+ echo "GSS_SK now at flavor: $flavor"
+ else
+ set_rule $FSNAME any any $flavor
+ wait_flavor all2all $flavor
+ fi
}
}
min_ost_size () {
- $LCTL get_param -n osc.*.kbytesavail | sort -n | head -n1
+ $LFS df | grep OST | awk '{print $4}' | sort -un | head -1
}
#
local parent=$(dirname $path)
[ -d $path ] && return 0
- [ ! -d ${parent} ] && mkdir -p ${parent}
+ if [ ! -d ${parent} ]; then
+ mkdir -p ${parent} ||
+ error "mkdir parent '$parent' failed"
+ fi
fi
if [ $MDSCOUNT -le 1 ]; then
- mkdir $path
+ mkdir $path || error "mkdir '$path' failed"
else
local test_num=$(echo $testnum | sed -e 's/[^0-9]*//g')
local mdt_index
mdt_index=$stripe_index
fi
echo "striped dir -i$mdt_index -c$stripe_count $path"
- $LFS setdirstripe -i$mdt_index -c$stripe_count $path
+ $LFS mkdir -i$mdt_index -c$stripe_count $path ||
+ error "mkdir -i $mdt_index -c$stripe_count $path failed"
fi
}
local param=$($LFS getstripe -d $1 | parse_layout_param)
echo "$param"
}
+
+lfsck_verify_pfid()
+{
+ local f
+ local rc=0
+
+ # Cancel locks before setting lfsck_verify_pfid so that errors are more
+ # controllable
+ cancel_lru_locks mdc
+ cancel_lru_locks osc
+
+ # make sure PFID is set correctly for files
+ do_nodes $(comma_list $(osts_nodes)) \
+ "$LCTL set_param -n obdfilter.${FSNAME}-OST*.lfsck_verify_pfid=1"
+
+ for f in "$@"; do
+ cat $f &> /dev/nullA ||
+ { rc=$?; echo "verify $f failed"; break; }
+ done
+
+ do_nodes $(comma_list $(osts_nodes)) \
+ "$LCTL set_param -n obdfilter.${FSNAME}-OST*.lfsck_verify_pfid=0"
+ return $rc
+}
+
+# check that clients "oscs" was evicted after "before"
+check_clients_evicted() {
+ local before=$1
+ shift
+ local oscs=${@}
+ local osc
+ local rc=0
+
+ for osc in $oscs; do
+ ((rc++))
+ echo "Check state for $osc"
+ local evicted=$(do_facet client $LCTL get_param osc.$osc.state |
+ tail -n 3 | awk -F"[ [,]" \
+ '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
+ if (($? == 0)) && (($evicted > $before)); then
+ echo "$osc is evicted at $evicted"
+ ((rc--))
+ fi
+ done
+
+ [ $rc -eq 0 ] || error "client not evicted from OST"
+}
+
+# check that clients OSCS current_state is FULL
+check_clients_full() {
+ local timeout=$1
+ shift
+ local oscs=${@}
+
+ for osc in $oscs; do
+ wait_update_facet client \
+ "lctl get_param -n osc.$osc.state |
+ grep 'current_state: FULL'" \
+ "current_state: FULL" $timeout
+ [ $? -eq 0 ] || error "$osc state is not FULL"
+ done
+}