#!/bin/bash
-# vim:expandtab:shiftwidth=4:softtabstop=4:tabstop=4:
+# -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*-
+# vim:shiftwidth=4:softtabstop=4:tabstop=4:
#
# Tests for multiple mount protection (MMP) feature.
#
return ${PIPESTATUS[0]}
}
+# Set the MMP block to 'fsck' state
+mark_mmp_block() {
+ local facet=$1
+ local device=$2
+
+ do_facet $facet "$LUSTRE/tests/mmp_mark.sh $device"
+ return ${PIPESTATUS[0]}
+}
+
# Reset the MMP block (if any) back to the clean state.
reset_mmp_block() {
local facet=$1
# Get MMP update interval (in seconds) from the Lustre server target.
get_mmp_update_interval() {
- local facet=$1
- local device=$2
- local interval
-
- interval=$(do_facet $facet "$DEBUGFS -c -R dump_mmp $device 2>/dev/null \
- | grep 'MMP Update Interval' | cut -d' ' -f4")
- [ -z "$interval" ] && interval=1
-
- echo $interval
+ local facet=$1
+ local device=$2
+ local interval
+
+ interval=$(do_facet $facet \
+ "$DEBUGFS -c -R dump_mmp $device 2>$TMP/mmp.debugfs.msg" |
+ awk 'tolower($0) ~ /update.interval/ { print $NF }')
+ [ -z "$interval" ] && interval=5 &&
+ do_facet $facet cat $TMP/mmp.debugfs.msg &&
+ echo "$facet:$device: assume update interval=$interval" 1>&2 ||
+ echo "$facet:$device: got actual update interval=$interval" 1>&2
+
+ echo $interval
}
# Get MMP check interval (in seconds) from the Lustre server target.
get_mmp_check_interval() {
- local facet=$1
- local device=$2
- local interval
+ local facet=$1
+ local device=$2
+ local interval
+
+ interval=$(do_facet $facet \
+ "$DEBUGFS -c -R dump_mmp $device 2>$TMP/mmp.debugfs.msg" |
+ awk 'tolower($0) ~ /check.interval/ { print $NF }')
+ [ -z "$interval" ] && interval=5 &&
+ do_facet $facet cat $TMP/mmp.debugfs.msg &&
+ echo "$facet:$device: assume check interval=$interval" 1>&2 ||
+ echo "$facet:$device: got actual check interval=$interval" 1>&2
+
+ echo $interval
+}
- interval=$(do_facet $facet "$DEBUGFS -c -R dump_mmp $device 2>/dev/null \
- | grep 'MMP Check Interval' | cut -d' ' -f4")
- [ -z "$interval" ] && interval=5
+# Adjust the MMP update interval (in seconds) on the Lustre server target.
+# Specifying an interval of 0 means to use the default interval.
+set_mmp_update_interval() {
+ local facet=$1
+ local device=$2
+ local interval=${3:-0}
- echo $interval
+ do_facet $facet "$TUNE2FS -E mmp_update_interval=$interval $device"
+ return ${PIPESTATUS[0]}
}
+I_ENABLED_MDS=0
+I_ENABLED_OSS=0
+
# Enable the MMP feature on the Lustre server targets.
mmp_init() {
- init_vars
-
- # The MMP feature is automatically enabled by mkfs.lustre for
- # new file system at format time if failover is being used.
- # Otherwise, the Lustre administrator has to manually enable
- # this feature when the file system is unmounted.
-
- local var=${MMP_MDS}failover_HOST
- if [ -z "${!var}" ]; then
- log "Failover is not used on MDS, enabling MMP manually..."
- enable_mmp $MMP_MDS $MMP_MDSDEV || \
- error "failed to enable MMP on $MMP_MDSDEV on $MMP_MDS"
- fi
-
- var=${MMP_OSS}failover_HOST
- if [ -z "${!var}" ]; then
- log "Failover is not used on OSS, enabling MMP manually..."
- enable_mmp $MMP_OSS $MMP_OSTDEV || \
- error "failed to enable MMP on $MMP_OSTDEV on $MMP_OSS"
- fi
-
- # check whether the MMP feature is enabled or not
- mmp_is_enabled $MMP_MDS $MMP_MDSDEV || \
- error "MMP was not enabled on $MMP_MDSDEV on $MMP_MDS"
-
- mmp_is_enabled $MMP_OSS $MMP_OSTDEV || \
- error "MMP was not enabled on $MMP_OSTDEV on $MMP_OSS"
+ init_vars
+
+ if [ $(facet_fstype $MMP_MDS) != ldiskfs ]; then
+ skip "Only applicable to ldiskfs-based MDTs"
+ exit
+ fi
+
+ if [ $(facet_fstype $MMP_OSS) != ldiskfs ]; then
+ skip "Only applicable to ldiskfs-based OSTs"
+ exit
+ fi
+
+ mmp_is_enabled $MMP_MDS $MMP_MDSDEV ||
+ {
+ log "MMP is not enabled on MDS, enabling it manually..."
+ enable_mmp $MMP_MDS $MMP_MDSDEV ||
+ error "failed to enable MMP on $MMP_MDSDEV on $MMP_MDS"
+ I_ENABLED_MDS=1
+ }
+
+ mmp_is_enabled $MMP_OSS $MMP_OSTDEV ||
+ {
+ log "MMP is not enabled on OSS, enabling it manually..."
+ enable_mmp $MMP_OSS $MMP_OSTDEV ||
+ error "failed to enable MMP on $MMP_OSTDEV on $MMP_OSS"
+ I_ENABLED_OSS=1
+ }
+
+ # check whether the MMP feature is enabled or not
+ mmp_is_enabled $MMP_MDS $MMP_MDSDEV ||
+ error "MMP was not enabled on $MMP_MDSDEV on $MMP_MDS"
+
+ mmp_is_enabled $MMP_OSS $MMP_OSTDEV ||
+ error "MMP was not enabled on $MMP_OSTDEV on $MMP_OSS"
}
# Disable the MMP feature on the Lustre server targets
-# which did not use failover.
mmp_fini() {
- local var=${MMP_MDS}failover_HOST
- if [ -z "${!var}" ]; then
- log "Failover is not used on MDS, disabling MMP manually..."
- disable_mmp $MMP_MDS $MMP_MDSDEV || \
- error "failed to disable MMP on $MMP_MDSDEV on $MMP_MDS"
- mmp_is_enabled $MMP_MDS $MMP_MDSDEV && \
- error "MMP was not disabled on $MMP_MDSDEV on $MMP_MDS"
- fi
-
- var=${MMP_OSS}failover_HOST
- if [ -z "${!var}" ]; then
- log "Failover is not used on OSS, disabling MMP manually..."
- disable_mmp $MMP_OSS $MMP_OSTDEV || \
- error "failed to disable MMP on $MMP_OSTDEV on $MMP_OSS"
- mmp_is_enabled $MMP_OSS $MMP_OSTDEV && \
- error "MMP was not disabled on $MMP_OSTDEV on $MMP_OSS"
- fi
-
- return 0
+ if [ $I_ENABLED_MDS -eq 1 ]; then
+ log "Disabling MMP on $MMP_MDSDEV on $MMP_MDS manually..."
+ disable_mmp $MMP_MDS $MMP_MDSDEV ||
+ error "failed to disable MMP on $MMP_MDSDEV on $MMP_MDS"
+ mmp_is_enabled $MMP_MDS $MMP_MDSDEV &&
+ error "MMP was not disabled on $MMP_MDSDEV on $MMP_MDS"
+ fi
+
+ if [ $I_ENABLED_OSS -eq 1 ]; then
+ log "Disabling MMP on $MMP_OSTDEV on $MMP_OSS manually..."
+ disable_mmp $MMP_OSS $MMP_OSTDEV ||
+ error "failed to disable MMP on $MMP_OSTDEV on $MMP_OSS"
+ mmp_is_enabled $MMP_OSS $MMP_OSTDEV &&
+ error "MMP was not disabled on $MMP_OSTDEV on $MMP_OSS"
+ fi
+
+ return 0
}
-# Mount the shared target on the failover server after some interval it's
+# Mount the shared target on the failover server after some interval it's
# mounted on the primary server.
mount_after_interval_sub() {
local interval=$1
return ${PIPESTATUS[0]}
return 1
elif [ $second_mount_rc -ne 0 -a $first_mount_rc -ne 0 ]; then
- error_noexit "failed to mount on the failover pair $facet,$failover_facet"
+ error_noexit "mount failure on failover pair $facet,$failover_facet"
return $first_mount_rc
fi
return 0
}
-# Mount the shared target on the failover server
+# Mount the shared target on the failover server
# during unmounting it on the primary server.
mount_during_unmount() {
local device=$1
return 0
}
-# Mount the shared target on the failover server
+# Mount the shared target on the failover server
# after clean unmounting it on the primary server.
mount_after_unmount() {
local device=$1
start $facet $device $mnt_opts || return ${PIPESTATUS[0]}
log "Unmounting $device on $facet..."
- stop $facet || return ${PIPESTATUS[0]}
+ stop $facet || return ${PIPESTATUS[0]}
log "Mounting $device on $failover_facet..."
start $failover_facet $device $mnt_opts || return ${PIPESTATUS[0]}
# Run e2fsck on the Lustre server target.
run_e2fsck() {
- local facet=$1
- shift
- local device=$1
- shift
- local opts="$@"
-
- log "Running e2fsck on the device $device on $facet..."
- do_facet $facet "$E2FSCK $opts $device"
- return ${PIPESTATUS[0]}
+ local facet=$1
+ shift
+ local device=$1
+ shift
+ local opts="$@"
+
+ echo "Running e2fsck on the device $device on $facet..."
+ do_facet $facet "$E2FSCK $opts $device"
+ return ${PIPESTATUS[0]}
}
# Check whether there are failover pairs for MDS and OSS servers.
# Test 8 - mount during e2fsck (should never succeed).
test_8() {
- local e2fsck_pid
-
- run_e2fsck $MMP_MDS $MMP_MDSDEV "-fy" &
- e2fsck_pid=$!
- sleep 1
-
- log "Mounting $MMP_MDSDEV on $MMP_MDS_FAILOVER..."
- if start $MMP_MDS_FAILOVER $MMP_MDSDEV $MDS_MOUNT_OPTS; then
- error_noexit "mount $MMP_MDSDEV on $MMP_MDS_FAILOVER should fail"
- stop $MMP_MDS_FAILOVER || return ${PIPESTATUS[0]}
- return 1
- fi
-
- wait $e2fsck_pid
-
- echo
- run_e2fsck $MMP_OSS $MMP_OSTDEV "-fy" &
- e2fsck_pid=$!
- sleep 1
-
- log "Mounting $MMP_OSTDEV on $MMP_OSS_FAILOVER..."
- if start $MMP_OSS_FAILOVER $MMP_OSTDEV $OST_MOUNT_OPTS; then
- error_noexit "mount $MMP_OSTDEV on $MMP_OSS_FAILOVER should fail"
- stop $MMP_OSS_FAILOVER || return ${PIPESTATUS[0]}
- return 2
- fi
-
- wait $e2fsck_pid
- return 0
+ local e2fsck_pid
+ local saved_interval
+ local new_interval
+
+ # After writing a new sequence number into the MMP block, e2fsck will
+ # sleep at least (2 * new_interval + 1) seconds before it goes into
+ # e2fsck passes.
+ new_interval=30
+
+ # MDT
+ saved_interval=$(get_mmp_update_interval $MMP_MDS $MMP_MDSDEV)
+ set_mmp_update_interval $MMP_MDS $MMP_MDSDEV $new_interval
+
+ run_e2fsck $MMP_MDS $MMP_MDSDEV "-fy" &
+ e2fsck_pid=$!
+ sleep 5
+
+ if start $MMP_MDS_FAILOVER $MMP_MDSDEV $MDS_MOUNT_OPTS; then
+ error_noexit \
+ "mount $MMP_MDSDEV on $MMP_MDS_FAILOVER should fail"
+ stop $MMP_MDS_FAILOVER || return ${PIPESTATUS[0]}
+ set_mmp_update_interval $MMP_MDS $MMP_MDSDEV $saved_interval
+ return 1
+ fi
+
+ wait $e2fsck_pid
+ set_mmp_update_interval $MMP_MDS $MMP_MDSDEV $saved_interval
+
+ # OST
+ echo
+ saved_interval=$(get_mmp_update_interval $MMP_OSS $MMP_OSTDEV)
+ set_mmp_update_interval $MMP_OSS $MMP_OSTDEV $new_interval
+
+ run_e2fsck $MMP_OSS $MMP_OSTDEV "-fy" &
+ e2fsck_pid=$!
+ sleep 5
+
+ if start $MMP_OSS_FAILOVER $MMP_OSTDEV $OST_MOUNT_OPTS; then
+ error_noexit \
+ "mount $MMP_OSTDEV on $MMP_OSS_FAILOVER should fail"
+ stop $MMP_OSS_FAILOVER || return ${PIPESTATUS[0]}
+ set_mmp_update_interval $MMP_OSS $MMP_OSTDEV $saved_interval
+ return 2
+ fi
+
+ wait $e2fsck_pid
+ set_mmp_update_interval $MMP_OSS $MMP_OSTDEV $saved_interval
+ return 0
}
run_test 8 "mount during e2fsck"
# Test 9 - mount after aborted e2fsck (should never succeed).
test_9() {
- local e2fsck_pid
- local mdt_mmp_check_interval
- local ost_mmp_check_interval
-
start $MMP_MDS $MMP_MDSDEV $MDS_MOUNT_OPTS || return ${PIPESTATUS[0]}
if ! start $MMP_OSS $MMP_OSTDEV $OST_MOUNT_OPTS; then
local rc=${PIPESTATUS[0]}
fi
stop_services primary || return ${PIPESTATUS[0]}
- mdt_mmp_check_interval=$(get_mmp_check_interval $MMP_MDS $MMP_MDSDEV)
- run_e2fsck $MMP_MDS_FAILOVER $MMP_MDSDEV "-fy" &
- e2fsck_pid=$!
- sleep $((2 * $mdt_mmp_check_interval + 1))
- kill -s ABRT $e2fsck_pid
+ mark_mmp_block $MMP_MDS $MMP_MDSDEV || return ${PIPESTATUS[0]}
log "Mounting $MMP_MDSDEV on $MMP_MDS..."
if start $MMP_MDS $MMP_MDSDEV $MDS_MOUNT_OPTS; then
reset_mmp_block $MMP_MDS $MMP_MDSDEV || return ${PIPESTATUS[0]}
- echo
- ost_mmp_check_interval=$(get_mmp_check_interval $MMP_OSS $MMP_OSTDEV)
- run_e2fsck $MMP_OSS_FAILOVER $MMP_OSTDEV "-fy" &
- e2fsck_pid=$!
- sleep $((2 * $ost_mmp_check_interval + 1))
- kill -s ABRT $e2fsck_pid
+ mark_mmp_block $MMP_OSS $MMP_OSTDEV || return ${PIPESTATUS[0]}
log "Mounting $MMP_OSTDEV on $MMP_OSS..."
if start $MMP_OSS $MMP_OSTDEV $OST_MOUNT_OPTS; then
run_e2fsck $MMP_MDS_FAILOVER $MMP_MDSDEV "-fn"
rc=${PIPESTATUS[0]}
- if [ $rc -ne 8 ]; then
- error_noexit "e2fsck $MMP_MDSDEV on $MMP_MDS_FAILOVER should return 8"
+
+ # e2fsck is called with -n option (Open the filesystem read-only), so
+ # 0 (No errors) and 4 (File system errors left uncorrected) are the only
+ # acceptable exit codes in this case
+ if [ $rc -ne 0 ] && [ $rc -ne 4 ]; then
+ error_noexit "e2fsck $MMP_MDSDEV on $MMP_MDS_FAILOVER returned $rc"
stop $MMP_MDS || return ${PIPESTATUS[0]}
- [ $rc -ne 0 ] && return $rc || return 1
+ return $rc
fi
log "Mounting $MMP_OSTDEV on $MMP_OSS..."
run_e2fsck $MMP_OSS_FAILOVER $MMP_OSTDEV "-fn"
rc=${PIPESTATUS[0]}
- if [ $rc -ne 8 ]; then
- error_noexit "e2fsck $MMP_OSTDEV on $MMP_OSS_FAILOVER should return 8"
- stop_services primary || return ${PIPESTATUS[0]}
- [ $rc -ne 0 ] && return $rc || return 1
+ if [ $rc -ne 0 ] && [ $rc -ne 4 ]; then
+ error_noexit "e2fsck $MMP_OSTDEV on $MMP_OSS_FAILOVER returned $rc"
fi
stop_services primary || return ${PIPESTATUS[0]}
mmp_fini
FAIL_ON_ERROR=$SAVED_FAIL_ON_ERROR
-equals_msg $(basename $0): test complete
+complete $SECONDS
$MMP_RESTORE_MOUNT && setupall
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && \
- grep -q FAIL $TESTSUITELOG && exit 1 || true
-echo "$0: completed"
+exit_status
git://git.whamcloud.com / fs / lustre-release.git / blobdiff