Whamcloud - gitweb
LU-13449 tests: fix recovery-small test_140b check
[fs/lustre-release.git] / lustre / tests / recovery-small.sh
index 60b78f2..8b4b594 100755 (executable)
@@ -9,11 +9,6 @@ init_test_env $@
 init_logging
 
 ALWAYS_EXCEPT="$RECOVERY_SMALL_EXCEPT "
-if $SHARED_KEY; then
-       # bug number for skipped test: LU-12896
-       ALWAYS_EXCEPT+="               110k"
-       # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
-fi
 
 build_test_filter
 
@@ -2971,11 +2966,12 @@ test_140b() {
        replay_barrier mds1
        umount_mds_client
        fail mds1
+       # Lustre: tfs-MDT0000: Recovery over after 0:03, of 2 clients 2 rec...
        local recovery=$(do_facet mds1 dmesg |
-                        awk -F: '/Recovery over after/ { print $4 }' |
-                        cut -d, -f1 | tail -1)
-       (( $recovery < $TIMEOUT*2 )) ||
-               error "recovery took too long $recovery > $((TIMEOUT * 2))"
+                        awk '/Recovery over after/ { print $6 }' | tail -1 |
+                        awk -F: '{ print $1 * 60 + $2 }')
+       (( recovery < TIMEOUT * 2 + 5 )) ||
+               error "recovery took too long $recovery > $((TIMEOUT * 2 + 5))"
 }
 run_test 140b "local mount is excluded from recovery"
 
@@ -3076,6 +3072,87 @@ $(do_facet mds1 $LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_time_hard)
 }
 run_test 145 "connect mdtlovs and process update logs after recovery expire"
 
+test_147() {
+       local obd_timeout=200
+       local old=$($LCTL get_param -n timeout)
+       local f=$DIR/$tfile
+       local connection_count
+
+       $LFS setstripe -i 0 -c 1 $f
+       stripe_index=$($LFS getstripe -i $f)
+       if [ $stripe_index -ne 0 ]; then
+              $LFS getstripe $f
+              error "$f: stripe_index $stripe_index != 0" && return
+       fi
+
+       $LCTL set_param timeout=$obd_timeout
+       stack_trap "$LCTL set_param timeout=$old && client_reconnect" EXIT
+
+       # OBD_FAIL_OST_CONNECT_NET2
+       # lost reply to connect request
+       do_facet ost1 lctl set_param fail_loc=0x00000225 timeout=$obd_timeout
+       stack_trap "do_facet ost1 $LCTL set_param fail_loc=0 timeout=$old" EXIT
+
+
+       ost_evict_client
+       # force reconnect
+       $LFS df $MOUNT > /dev/null 2>&1 &
+       sleep $((obd_timeout * 3 / 4))
+
+       $LCTL get_param osc.$FSNAME-OST0000-osc-*.state
+       connection_count=$($LCTL get_param osc.$FSNAME-OST0000-osc-*.state |
+                          tac |  sed "/FULL/,$ d" | grep CONNECTING | wc -l)
+
+       echo $connection_count
+       (($connection_count >= 6)) || error "Client reconnected too slow"
+}
+run_test 147 "Check client reconnect"
+
+test_148() {
+       local wce_param="obdfilter.$FSNAME-OST0000.writethrough_cache_enable"
+       local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
+       local amc=$(at_max_get client)
+       local amo=$(at_max_get ost1)
+       local timeout
+
+       at_max_set 0 client
+       at_max_set 0 ost1
+       timeout=$(request_timeout client)
+
+       [ "$(facet_fstype ost1)" = "ldiskfs" ] && {
+               # save old r/o cache settings
+               save_lustre_params ost1 $wce_param > $p
+
+               # disable r/o cache
+               do_facet ost1 "$LCTL set_param -n $wce_param=0"
+       }
+
+       $LFS setstripe -i 0 -c 1 $DIR/$tfile
+       dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=direct
+       cp $DIR/$tfile $TMP/$tfile
+       #define OBD_FAIL_OST_BRW_PAUSE_BULK2     0x227
+       do_facet ost1 $LCTL set_param fail_loc=0x80000227
+       do_facet ost1 $LCTL set_param fail_val=$((timeout+2))
+       dd if=/dev/urandom of=$DIR/$tfile bs=4096 count=1 conv=notrunc,fdatasync
+       dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=notrunc,fdatasync
+       sleep 2
+       cancel_lru_locks osc
+       cmp -b $DIR/$tfile $TMP/$tfile || error "wrong data"
+
+       rm -f $DIR/$tfile $TMP/$tfile
+
+       at_max_set $amc client
+       at_max_set $amo ost1
+
+       [ "$(facet_fstype ost1)" = "ldiskfs" ] && {
+               # restore initial r/o cache settings
+               restore_lustre_params < $p
+       }
+
+       return 0
+}
+run_test 148 "data corruption through resend"
+
 complete $SECONDS
 check_and_cleanup_lustre
 exit_status