LU-482 tests: cleanup clients/OSTs after MDS error

[fs/lustre-release.git] / lustre / tests / insanity.sh

diff --git a/lustre/tests/insanity.sh b/lustre/tests/insanity.sh
index ebf5cb8..f0cb403 100755 (executable)
--- a/lustre/tests/insanity.sh
+++ b/lustre/tests/insanity.sh
@@ -9,14 +9,13 @@ LUSTRE=${LUSTRE:-`dirname $0`/..}
 init_test_env $@
 
 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+init_logging
 #              
 ALWAYS_EXCEPT="10  $INSANITY_EXCEPT"
 
 if [ "$FAILURE_MODE" = "HARD" ]; then
-    mixed_ost_devs && CONFIG_EXCEPTIONS="0 2 4 5 6 8" && \
-        echo -n "Several ost services on one ost node are used with FAILURE_MODE=$FAILURE_MODE. " && \
-        echo "Except the tests: $CONFIG_EXCEPTIONS" && \
-        ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
+        skip_env "$0: is not functional with FAILURE_MODE = HARD, please use recovery-double-scale, bz20407"
+        exit 0
 fi
 
 #
@@ -35,13 +34,14 @@ assert_env mds_HOST MDS_MKFS_OPTS
 assert_env ost_HOST OST_MKFS_OPTS OSTCOUNT
 assert_env LIVE_CLIENT FSNAME
 
-remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
-remote_ost_nodsh && skip "remote OST with nodsh" && exit 0
+require_dsh_mds || exit 0
+require_dsh_ost || exit 0
 
 # 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
@@ -60,19 +60,6 @@ set_fail_client() {
     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
-}
-
 fail_clients() {
     num=$1
 
@@ -124,29 +111,28 @@ 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
+       $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
+       $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" 
+       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"
+       echo "rmdir $TESTDIR/$c"
+       $PDSH $LIVE_CLIENT "rmdir $TESTDIR/$c"
     done
 }
 
@@ -157,18 +143,17 @@ clients_recover_osts() {
 
 check_and_setup_lustre
 
+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
 
     for i in $(seq $OSTCOUNT) ; do
-        facet_failover ost$i || return 4
-        echo "Waiting for df pid: $DFPID"
-        wait $DFPID || { echo "df returned $?" && return 3; }
+        fail ost$i
     done
     return 0
 }
@@ -187,7 +172,7 @@ test_2() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
 
-    client_df
+    clients_up
 
     shutdown_facet $SINGLEMDS
     reboot_facet $SINGLEMDS
@@ -196,7 +181,7 @@ test_2() {
     change_active $SINGLEMDS
     reboot_facet $SINGLEMDS
 
-    client_df &
+    clients_up &
     DFPID=$!
     sleep 5
 
@@ -204,17 +189,17 @@ test_2() {
 
     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`"
@@ -228,12 +213,11 @@ test_3() {
     [ -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"
@@ -241,13 +225,13 @@ test_3() {
     
     #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`"
@@ -262,7 +246,7 @@ test_4() {
  
     #Check FS
     echo "Test Lustre stability after OST failure"
-    client_df &
+    clients_up &
     DFPIDA=$!
     sleep 5
 
@@ -274,17 +258,17 @@ test_4() {
     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
     
@@ -292,14 +276,14 @@ test_4() {
     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() {
-    [ $OSTCOUNT -lt 2 ] && skip "$OSTCOUNT < 2, not enough OSTs" && return 0
+    [ $OSTCOUNT -lt 2 ] && skip_env "$OSTCOUNT < 2, not enough OSTs" && return 0
 
     echo "Fifth Failure Mode: OST/OST `date`"
 
@@ -307,7 +291,7 @@ test_5() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
 
-    client_df
+    clients_up
     
     #OST Portion
     shutdown_facet ost1
@@ -315,7 +299,7 @@ test_5() {
     
     #Check FS
     echo "Test Lustre stability after OST failure"
-    client_df &
+    clients_up &
     DFPIDA=$!
     sleep 5
     
@@ -325,15 +309,15 @@ test_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
@@ -342,7 +326,7 @@ test_5() {
 
     wait $DFPIDA
     wait $DFPIDB
-    client_df || return 2
+    clients_up || return 2
 }
 run_test 5 "Fifth Failure Mode: OST/OST `date`"
 ###################################################
@@ -355,7 +339,7 @@ test_6() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
 
-    client_df
+    clients_up
     client_touch testfile || return 2
        
     #OST Portion
@@ -364,7 +348,7 @@ test_6() {
 
     #Check FS
     echo "Test Lustre stability after OST failure"
-    client_df &
+    clients_up &
     DFPIDA=$!
     echo DFPIDA=$DFPIDA
     sleep 5
@@ -375,25 +359,25 @@ test_6() {
     
     #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 || return 1
     sleep 5 
 
-    wait_remote_prog df $((TIMEOUT * 3 + 10)) 
+    wait_remote_prog "stat -f" $((TIMEOUT * 3 + 20)) 
     wait $DFPIDA
     wait $DFPIDB
 
     echo "Verifying mount"
     [ -z "$(mounted_lustre_filesystems)" ] && return 3
-    client_df
+    clients_up
 }
 run_test 6 "Sixth Failure Mode: OST/CLIENT `date`"
 ###################################################
@@ -407,7 +391,7 @@ test_7() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
 
-    client_df
+    clients_up
     client_touch testfile  || return 1
 
     #CLIENT Portion
@@ -416,9 +400,9 @@ test_7() {
     
     #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"
@@ -428,22 +412,19 @@ test_7() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && return 2
 
-    client_df
+    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 || return 2
-    client_df
+    clients_up
     
     #Sleep
     echo "wait 1 minutes"
@@ -461,7 +442,7 @@ test_8() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
 
-    client_df
+    clients_up
     client_touch testfile
        
     #CLIENT Portion
@@ -470,9 +451,9 @@ test_8() {
 
     #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"
@@ -482,7 +463,7 @@ test_8() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
 
-    client_df
+    clients_up
     client_touch testfile
 
 
@@ -492,20 +473,20 @@ test_8() {
 
     #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 || return 3
-    wait_for ost1
+    wait_for_facet ost1
     start_ost 1
     wait $DFPID
-    client_df || return 1
+    clients_up || return 1
     client_touch testfile2 || return 2
 
     #Sleep
@@ -524,7 +505,7 @@ test_9() {
     echo "Verify Lustre filesystem is up and running"
     [ -z "$(mounted_lustre_filesystems)" ] && error "Lustre is not running"
 
-    client_df
+    clients_up
     client_touch testfile || return 1
        
     #CLIENT Portion
@@ -533,9 +514,9 @@ test_9() {
 
     #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"
@@ -543,7 +524,7 @@ test_9() {
 
     #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
@@ -552,14 +533,14 @@ test_9() {
     
     #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 || return 7
-    client_df
+    clients_up
     
     #Sleep
     echo "Wait 1 minutes"
@@ -576,6 +557,6 @@ test_10() {
 }
 run_test 10 "Running Availability for 6 hours..."
 
-equals_msg `basename $0`: test complete, cleaning up
+complete $(basename $0) $SECONDS
 check_and_cleanup_lustre
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true
+exit_status