LU-1415 tests: Adapt oos to the new grant and osd-zfs behavior

This was originally developed by Johann Lombardi under ORI-151 and
ORI-398.  See http://review.whamcloud.com/1545 and
http://review.whamcloud.com/1665.

Change-Id: If778df1d672e2adb182cb801036aaef950d80af4
Whamcloud-bug-id: ORI-151
Whamcloud-bug-id: ORI-398
Signed-off-by: Li Wei <liwei@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/2978
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Reviewed-by: Niu Yawei <niu@whamcloud.com>

diff --git a/lustre/tests/oos.sh b/lustre/tests/oos.sh
index c547b63..5d6a161 100755 (executable)
--- a/lustre/tests/oos.sh
+++ b/lustre/tests/oos.sh
@@ -46,7 +46,11 @@ oos_full || true
 
 # make sure we stripe over all OSTs to avoid OOS on only a subset of OSTs
 $LFS setstripe $OOS -c $STRIPECOUNT
-if dd if=/dev/zero of=$OOS count=$(($ORIGFREE + 100)) bs=1k 2> $LOG; then
+# add 20% of margin since the metadata overhead estimated in bavail might be
+# too aggressive and we might be able to write more than reported initially
+# by statfs.
+echo dd size $((ORIGFREE * 120 / 100))kB
+if dd if=/dev/zero of=$OOS count=$((ORIGFREE * 120 / 100)) bs=1k 2> $LOG; then
        echo "ERROR: dd did not fail"
        SUCCESS=0
 fi
@@ -81,10 +85,8 @@ fi
 
 #$LCTL debug_daemon stop
 
-echo LOG file
-cat $LOG
+[ $SUCCESS != 0 ] && echo LOG file && sed "s/^/LOG: /" $LOG
 rm -f $OOS
-sync; sleep 1; sync
 
 sync; sleep 3; sync
 

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index fe080f5..a7c660a 100644 (file)
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -4566,20 +4566,30 @@ wait_clients_import_state () {
 }
 
 oos_full() {
-        local -a AVAILA
-        local -a GRANTA
-        local OSCFULL=1
-        AVAILA=($(do_nodes $(comma_list $(osts_nodes)) \
-                  $LCTL get_param obdfilter.*.kbytesavail))
-        GRANTA=($(do_nodes $(comma_list $(osts_nodes)) \
-                  $LCTL get_param -n obdfilter.*.tot_granted))
-        for ((i=0; i<${#AVAILA[@]}; i++)); do
-                local -a AVAIL1=(${AVAILA[$i]//=/ })
-                GRANT=$((${GRANTA[$i]}/1024))
-                echo -n $(echo ${AVAIL1[0]} | cut -d"." -f2) avl=${AVAIL1[1]} grnt=$GRANT diff=$((AVAIL1[1] - GRANT))
-                [ $((AVAIL1[1] - GRANT)) -lt 400 ] && OSCFULL=0 && echo " FULL" || echo
-        done
-        return $OSCFULL
+       local -a AVAILA
+       local -a GRANTA
+       local -a TOTALA
+       local OSCFULL=1
+       AVAILA=($(do_nodes $(comma_list $(osts_nodes)) \
+                 $LCTL get_param obdfilter.*.kbytesavail))
+       GRANTA=($(do_nodes $(comma_list $(osts_nodes)) \
+                 $LCTL get_param -n obdfilter.*.tot_granted))
+       TOTALA=($(do_nodes $(comma_list $(osts_nodes)) \
+                 $LCTL get_param -n obdfilter.*.kbytestotal))
+       for ((i=0; i<${#AVAILA[@]}; i++)); do
+               local -a AVAIL1=(${AVAILA[$i]//=/ })
+               local -a TOTAL=(${TOTALA[$i]//=/ })
+               GRANT=$((${GRANTA[$i]}/1024))
+               # allow 1% of total space in bavail because of delayed
+               # allocation with ZFS which might release some free space after
+               # txg commit.  For small devices, we set a mininum of 8MB
+               local LIMIT=$((${TOTAL} / 100 + 8000))
+               echo -n $(echo ${AVAIL1[0]} | cut -d"." -f2) avl=${AVAIL1[1]} \
+                       grnt=$GRANT diff=$((AVAIL1[1] - GRANT)) limit=${LIMIT}
+               [ $((AVAIL1[1] - GRANT)) -lt $LIMIT ] && OSCFULL=0 && \
+                       echo " FULL" || echo
+       done
+       return $OSCFULL
 }
 
 pool_list () {