-#!/bin/sh
+#!/bin/bash
+# -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*-
+# vim:shiftwidth=4:softtabstop=4:tabstop=4:
+#
# Test multiple failures, AKA Test 17
set -e
init_test_env $@
-. ${CONFIG:=$LUSTRE/tests/cfg/insanity-local.sh}
+. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+init_logging
+#
+ALWAYS_EXCEPT="10 $INSANITY_EXCEPT"
-ALWAYS_EXCEPT="10 $INSANITY_EXCEPT"
+if [ "$FAILURE_MODE" = "HARD" ]; then
+ skip_env "$0: is not functional with FAILURE_MODE = HARD, please use recovery-double-scale, bz20407"
+ exit 0
+fi
+
+#
+[ "$SLOW" = "no" ] && EXCEPT_SLOW=""
-SETUP=${SETUP:-"setup"}
-CLEANUP=${CLEANUP:-"cleanup"}
+SETUP=${SETUP:-""}
+CLEANUP=${CLEANUP:-""}
build_test_filter
-assert_env mds_HOST MDS_MKFS_OPTS
-assert_env ost_HOST OST_MKFS_OPTS OSTCOUNT
+SINGLECLIENT=${SINGLECLIENT:-$HOSTNAME}
+LIVE_CLIENT=${LIVE_CLIENT:-$SINGLECLIENT}
+FAIL_CLIENTS=${FAIL_CLIENTS:-$RCLIENTS}
+
+assert_env mds_HOST
+assert_env ost_HOST OSTCOUNT
assert_env LIVE_CLIENT FSNAME
+require_dsh_mds || exit 0
+require_dsh_ost || exit 0
-# This can be a regexp, to allow more clients
-CLIENTS=${CLIENTS:-"`comma_list $LIVE_CLIENT $FAIL_CLIENTS $EXTRA_CLIENTS`"}
+# FAIL_CLIENTS list should not contain the LIVE_CLIENT
+FAIL_CLIENTS=$(echo " $FAIL_CLIENTS " | sed -re "s/\s+$LIVE_CLIENT\s+/ /g")
DIR=${DIR:-$MOUNT}
+TESTDIR=$DIR/d0.$(basename $0 .sh)
#####
# fail clients round robin
echo "fail $FAIL_CLIENT, next is $FAIL_NEXT"
}
-shutdown_client() {
- client=$1
- if [ "$FAILURE_MODE" = HARD ]; then
- $POWER_DOWN $client
- while ping -w 3 -c 1 $client > /dev/null 2>&1; do
- echo "waiting for node $client to fail"
- sleep 1
- done
- elif [ "$FAILURE_MODE" = SOFT ]; then
- zconf_umount $client $MOUNT -f
- fi
-}
-
-reboot_node() {
- NODE=$1
- if [ "$FAILURE_MODE" = HARD ]; then
- $POWER_UP $NODE
- fi
-}
-
fail_clients() {
- num=$1
- if [ -z "$num" ] || [ "$num" -gt $((FAIL_NUM - DOWN_NUM)) ]; then
- num=$((FAIL_NUM - DOWN_NUM))
- fi
+ num=$1
+
+ log "Request fail clients: $num, to fail: $FAIL_NUM, failed: $DOWN_NUM"
+ if [ -z "$num" ] || [ "$num" -gt $((FAIL_NUM - DOWN_NUM)) ]; then
+ num=$((FAIL_NUM - DOWN_NUM))
+ fi
if [ -z "$num" ] || [ "$num" -le 0 ]; then
+ log "No clients failed!"
return
fi
echo "down clients: $DOWN_CLIENTS"
- for client in $DOWN_CLIENTS; do
- reboot_node $client
- done
- DOWN_NUM=`echo $DOWN_CLIENTS | wc -w`
- client_rmdirs
+ for client in $DOWN_CLIENTS; do
+ boot_node $client
+ done
+ DOWN_NUM=`echo $DOWN_CLIENTS | wc -w`
+ client_rmdirs
}
reintegrate_clients() {
- for client in $DOWN_CLIENTS; do
- wait_for_host $client
- echo "Restarting $client"
- zconf_mount $client $MOUNT || return 1
- done
- DOWN_CLIENTS=""
- DOWN_NUM=0
+ for client in $DOWN_CLIENTS; do
+ wait_for_host $client
+ echo "Restarting $client"
+ zconf_mount $client $MOUNT || return 1
+ done
+
+ DOWN_CLIENTS=""
+ DOWN_NUM=0
}
start_ost() {
- start ost$1 `ostdevname $1` $OST_MOUNT_OPTS
-}
-
-setup() {
- cleanup
- rm -rf logs/*
- formatall
- setupall
-
- while ! do_node $CLIENTS "ls -d $LUSTRE" > /dev/null; do sleep 5; done
- grep " $MOUNT " /proc/mounts || zconf_mount $CLIENTS $MOUNT
-}
-
-cleanup() {
- zconf_umount $CLIENTS $MOUNT
- cleanupall
- cleanup_check
+ start ost$1 `ostdevname $1` $OST_MOUNT_OPTS
}
trap exit INT
client_touch() {
- file=$1
- for c in $LIVE_CLIENT $FAIL_CLIENTS; do
- if echo $DOWN_CLIENTS | grep -q $c; then continue; fi
- $PDSH $c touch $MOUNT/${c}_$file || return 1
- done
+ file=$1
+ for c in $LIVE_CLIENT $FAIL_CLIENTS; do
+ echo $DOWN_CLIENTS | grep -q $c && continue
+ $PDSH $c touch $TESTDIR/${c}_$file || return 1
+ done
}
client_rm() {
- file=$1
- for c in $LIVE_CLIENT $FAIL_CLIENTS; do
- $PDSH $c rm $MOUNT/${c}_$file
- done
+ file=$1
+ for c in $LIVE_CLIENT $FAIL_CLIENTS; do
+ $PDSH $c rm $TESTDIR/${c}_$file
+ done
}
client_mkdirs() {
- for c in $LIVE_CLIENT $FAIL_CLIENTS; do
- echo "$c mkdir $MOUNT/$c"
- $PDSH $c "mkdir $MOUNT/$c"
- $PDSH $c "ls -l $MOUNT/$c"
- done
+ for c in $LIVE_CLIENT $FAIL_CLIENTS; do
+ echo "$c mkdir $TESTDIR/$c"
+ $PDSH $c "mkdir $TESTDIR/$c && ls -l $TESTDIR/$c"
+ done
}
client_rmdirs() {
- for c in $LIVE_CLIENT $FAIL_CLIENTS; do
- echo "rmdir $MOUNT/$c"
- $PDSH $LIVE_CLIENT "rmdir $MOUNT/$c"
- done
+ for c in $LIVE_CLIENT $FAIL_CLIENTS; do
+ echo "rmdir $TESTDIR/$c"
+ $PDSH $LIVE_CLIENT "rmdir $TESTDIR/$c"
+ done
}
clients_recover_osts() {
# do_node $CLIENTS "$LCTL "'--device %OSC_`hostname`_'"${facet}_svc_MNT_client_facet recover"
}
-if [ "$ONLY" == "cleanup" ]; then
- $CLEANUP
- exit
-fi
+check_and_setup_lustre
-if [ ! -z "$EVAL" ]; then
- eval "$EVAL"
- exit $?
-fi
-
-$SETUP
-
-if [ "$ONLY" == "setup" ]; then
- exit 0
-fi
+rm -rf $TESTDIR
+mkdir -p $TESTDIR
# 9 Different Failure Modes Combinations
echo "Starting Test 17 at `date`"
test_0() {
- facet_failover $SINGLEMDS
- echo "Waiting for df pid: $DFPID"
- wait $DFPID || { echo "df returned $?" && return 1; }
+ fail $SINGLEMDS
- facet_failover ost1 || return 4
- echo "Waiting for df pid: $DFPID"
- wait $DFPID || { echo "df returned $?" && return 2; }
-
- facet_failover ost2 || return 5
- echo "Waiting for df pid: $DFPID"
- wait $DFPID || { echo "df returned $?" && return 3; }
+ for i in $(seq $OSTCOUNT) ; do
+ fail ost$i
+ done
return 0
}
run_test 0 "Fail all nodes, independently"
############### Second Failure Mode ###############
test_2() {
- echo "Verify Lustre filesystem is up and running"
- client_df
+ echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ [ "$(facet_fstype ost1)" = "zfs" ] &&
+ skip "LU-2059: no local config for ZFS OSTs" && return
+
+ clients_up
shutdown_facet $SINGLEMDS
reboot_facet $SINGLEMDS
change_active $SINGLEMDS
reboot_facet $SINGLEMDS
- client_df &
+ clients_up &
DFPID=$!
sleep 5
echo "Reintegrating OST"
reboot_facet ost1
- wait_for ost1
+ wait_for_facet ost1
start_ost 1 || return 2
- wait_for $SINGLEMDS
+ wait_for_facet $SINGLEMDS
start $SINGLEMDS `mdsdevname 1` $MDS_MOUNT_OPTS || return $?
#Check FS
wait $DFPID
clients_recover_osts ost1
echo "Verify reintegration"
- client_df || return 1
+ clients_up || return 1
}
run_test 2 "Second Failure Mode: MDS/OST `date`"
test_3() {
#Create files
echo "Verify Lustre filesystem is up and running"
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
#MDS Portion
- facet_failover $SINGLEMDS
- wait $DFPID || echo df failed: $?
+ fail $SINGLEMDS
#Check FS
echo "Test Lustre stability after MDS failover"
- client_df
+ clients_up
#CLIENT Portion
echo "Failing 2 CLIENTS"
#Check FS
echo "Test Lustre stability after CLIENT failure"
- client_df
+ clients_up
#Reintegration
echo "Reintegrating CLIENTS"
reintegrate_clients || return 1
- client_df || return 3
+ clients_up || return 3
+ sleep 2 # give it a little time for fully recovered before next test
}
run_test 3 "Thirdb Failure Mode: MDS/CLIENT `date`"
###################################################
############### Fourth Failure Mode ###############
test_4() {
- echo "Fourth Failure Mode: OST/MDS `date`"
+ echo "Fourth Failure Mode: OST/MDS `date`"
+
+ [ "$(facet_fstype ost1)" = "zfs" ] &&
+ skip "LU-2059: no local config for ZFS OSTs" && return
#OST Portion
shutdown_facet ost1
#Check FS
echo "Test Lustre stability after OST failure"
- client_df &
+ clients_up &
DFPIDA=$!
sleep 5
change_active $SINGLEMDS
reboot_facet $SINGLEMDS
- client_df &
+ clients_up &
DFPIDB=$!
sleep 5
#Reintegration
echo "Reintegrating OST"
reboot_facet ost1
- wait_for ost1
+ wait_for_facet ost1
start_ost 1
-
- wait_for $SINGLEMDS
+
+ wait_for_facet $SINGLEMDS
start $SINGLEMDS `mdsdevname 1` $MDS_MOUNT_OPTS
#Check FS
wait $DFPIDB
clients_recover_osts ost1
echo "Test Lustre stability after MDS failover"
- client_df || return 1
+ clients_up || return 1
}
run_test 4 "Fourth Failure Mode: OST/MDS `date`"
###################################################
############### Fifth Failure Mode ###############
test_5() {
+ [ $OSTCOUNT -lt 2 ] && skip_env "$OSTCOUNT < 2, not enough OSTs" && return 0
+
echo "Fifth Failure Mode: OST/OST `date`"
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ clients_up
#OST Portion
shutdown_facet ost1
#Check FS
echo "Test Lustre stability after OST failure"
- client_df &
+ clients_up &
DFPIDA=$!
sleep 5
#Check FS
echo "Test Lustre stability after OST failure"
- client_df &
+ clients_up &
DFPIDB=$!
sleep 5
#Reintegration
echo "Reintegrating OSTs"
- wait_for ost1
+ wait_for_facet ost1
start_ost 1
- wait_for ost2
+ wait_for_facet ost2
start_ost 2
clients_recover_osts ost1
wait $DFPIDA
wait $DFPIDB
- client_df || return 2
+ clients_up || return 2
}
run_test 5 "Fifth Failure Mode: OST/OST `date`"
###################################################
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df || return 1
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ clients_up
client_touch testfile || return 2
#OST Portion
#Check FS
echo "Test Lustre stability after OST failure"
- client_df &
+ clients_up &
DFPIDA=$!
+ echo DFPIDA=$DFPIDA
sleep 5
#CLIENT Portion
#Check FS
echo "Test Lustre stability after CLIENTs failure"
- client_df &
+ clients_up &
DFPIDB=$!
+ echo DFPIDB=$DFPIDB
sleep 5
#Reintegration
echo "Reintegrating OST/CLIENTs"
- wait_for ost1
+ wait_for_facet ost1
start_ost 1
- reintegrate_clients
+ reintegrate_clients || return 1
sleep 5
+ wait_remote_prog "stat -f" $((TIMEOUT * 3 + 20))
wait $DFPIDA
wait $DFPIDB
+
echo "Verifying mount"
- client_df || return 3
+ [ -z "$(mounted_lustre_filesystems)" ] && return 3
+ clients_up
}
run_test 6 "Sixth Failure Mode: OST/CLIENT `date`"
###################################################
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ clients_up
client_touch testfile || return 1
#CLIENT Portion
#Check FS
echo "Test Lustre stability after CLIENTs failure"
- client_df
- $PDSH $LIVE_CLIENT "ls -l $MOUNT"
- $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
+ clients_up
+ $PDSH $LIVE_CLIENT "ls -l $TESTDIR"
+ $PDSH $LIVE_CLIENT "rm -f $TESTDIR/*_testfile"
#Sleep
echo "Wait 1 minutes"
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df
+ [ -z "$(mounted_lustre_filesystems)" ] && return 2
+
+ clients_up
client_rm testfile
#MDS Portion
- facet_failover $SINGLEMDS
+ fail $SINGLEMDS
- #Check FS
- echo "Test Lustre stability after MDS failover"
- wait $DFPID || echo "df on down clients fails " || return 1
- $PDSH $LIVE_CLIENT "ls -l $MOUNT"
- $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
+ $PDSH $LIVE_CLIENT "ls -l $TESTDIR"
+ $PDSH $LIVE_CLIENT "rm -f $TESTDIR/*_testfile"
#Reintegration
echo "Reintegrating CLIENTs"
- reintegrate_clients
- client_df || return 2
+ reintegrate_clients || return 2
+ clients_up
#Sleep
echo "wait 1 minutes"
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ clients_up
client_touch testfile
#CLIENT Portion
#Check FS
echo "Test Lustre stability after CLIENTs failure"
- client_df
- $PDSH $LIVE_CLIENT "ls -l $MOUNT"
- $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
+ clients_up
+ $PDSH $LIVE_CLIENT "ls -l $TESTDIR"
+ $PDSH $LIVE_CLIENT "rm -f $TESTDIR/*_testfile"
#Sleep
echo "Wait 1 minutes"
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ clients_up
client_touch testfile
#Check FS
echo "Test Lustre stability after OST failure"
- client_df &
+ clients_up &
DFPID=$!
sleep 5
#non-failout hangs forever here
- #$PDSH $LIVE_CLIENT "ls -l $MOUNT"
- #$PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile"
+ #$PDSH $LIVE_CLIENT "ls -l $TESTDIR"
+ #$PDSH $LIVE_CLIENT "rm -f $TESTDIR/*_testfile"
#Reintegration
echo "Reintegrating CLIENTs/OST"
- reintegrate_clients
- wait_for ost1
+ reintegrate_clients || return 3
+ wait_for_facet ost1
start_ost 1
wait $DFPID
- client_df || return 1
+ clients_up || return 1
client_touch testfile2 || return 2
#Sleep
#Create files
echo "Verify Lustre filesystem is up and running"
- client_df
+ [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
+
+ clients_up
client_touch testfile || return 1
#CLIENT Portion
#Check FS
echo "Test Lustre stability after CLIENTs failure"
- client_df
- $PDSH $LIVE_CLIENT "ls -l $MOUNT" || return 1
- $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" || return 2
+ clients_up
+ $PDSH $LIVE_CLIENT "ls -l $TESTDIR" || return 1
+ $PDSH $LIVE_CLIENT "rm -f $TESTDIR/*_testfile" || return 2
#Sleep
echo "Wait 1 minutes"
#Create files
echo "Verify Lustre filesystem is up and running"
- $PDSH $LIVE_CLIENT df $MOUNT || return 3
+ client_up $LIVE_CLIENT || return 3
client_touch testfile || return 4
#CLIENT Portion
#Check FS
echo "Test Lustre stability after CLIENTs failure"
- client_df
- $PDSH $LIVE_CLIENT "ls -l $MOUNT" || return 5
- $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" || return 6
+ clients_up
+ $PDSH $LIVE_CLIENT "ls -l $TESTDIR" || return 5
+ $PDSH $LIVE_CLIENT "rm -f $TESTDIR/*_testfile" || return 6
#Reintegration
echo "Reintegrating CLIENTs/CLIENTs"
- reintegrate_clients
- client_df || return 7
+ reintegrate_clients || return 7
+ clients_up
#Sleep
echo "Wait 1 minutes"
}
run_test 10 "Running Availability for 6 hours..."
-equals_msg "Done, cleaning up"
-$CLEANUP
+complete $(basename $0) $SECONDS
+check_and_cleanup_lustre
+exit_status
git://git.whamcloud.com / fs / lustre-release.git / blobdiff