X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Foos.sh;h=ba9bec3b885688d05cfd6384cb16c6d027c4abab;hb=3667acae1dc5af1973734c09f341dabe0e14ccd4;hp=910346ed3f84c857bd7f03f4a107ba69e4cc38f3;hpb=cd9c585e8c7bdd6cfd802be64ef277dfd466be17;p=fs%2Flustre-release.git

diff --git a/lustre/tests/oos.sh b/lustre/tests/oos.sh
index 910346e..ba9bec3 100755
--- a/lustre/tests/oos.sh
+++ b/lustre/tests/oos.sh
@@ -1,7 +1,7 @@
 #!/bin/bash
 
 set -e
-set -vx
+#set -vx
 
 export PATH=`dirname $0`/../utils:$PATH
 LFS=${LFS:-lfs}
@@ -13,74 +13,80 @@ LOG=$TMP/ooslog
 
 SUCCESS=1
 
-rm -f $OOS
+rm -f $OOS $LOG
 
-sleep 1	# to ensure we get up-to-date statfs info
+sync; sleep 1; sync	# to ensure we get up-to-date statfs info
 
-#echo -1 > /proc/sys/portals/debug
-#echo 0x40a8 > /proc/sys/portals/subsystem_debug
+#echo -1 > /proc/sys/lnet/debug
+#echo 0x40a8 > /proc/sys/lnet/subsystem_debug
 #lctl clear
 #lctl debug_daemon start /r/tmp/debug 1024
 
 STRIPECOUNT=`cat /proc/fs/lustre/lov/*/activeobd | head -n 1`
 ORIGFREE=`cat /proc/fs/lustre/llite/*/kbytesavail | head -n 1`
-MAXFREE=${MAXFREE:-$((200000 * $STRIPECOUNT))}
+MAXFREE=${MAXFREE:-$((400000 * $STRIPECOUNT))}
 if [ $ORIGFREE -gt $MAXFREE ]; then
 	echo "skipping out-of-space test on $OSC"
-	echo "reports ${ORIGFREE}kB free, more tham MAXFREE ${MAXFREE}kB"
+	echo "reports ${ORIGFREE}kB free, more than MAXFREE ${MAXFREE}kB"
 	echo "increase $MAXFREE (or reduce test fs size) to proceed"
 	exit 0
 fi
 
 export LANG=C LC_LANG=C # for "No space left on device" message
 
-# make sure, that log file will be removed. Somehow it was possible 
-# to me, that log file had +a and could not be rewritten, what led
-# to test fail.
-chattr -ai $LOG >/dev/null 2>&1
-rm -f $LOG >/dev/null 2>&1
+[ -f $LOG ] && echo "ERROR: log file wasn't removed?" && exit 1
 
 # make sure we stripe over all OSTs to avoid OOS on only a subset of OSTs
-$LFS setstripe $OOS 65536 0 $STRIPECOUNT
+$LFS setstripe $OOS -c $STRIPECOUNT
 if dd if=/dev/zero of=$OOS count=$(($ORIGFREE + 100)) bs=1k 2> $LOG; then
 	echo "ERROR: dd did not fail"
 	SUCCESS=0
 fi
 
 if [ "`grep -c 'No space left on device' $LOG`" -ne 1 ]; then
-        echo "ERROR: dd not return ENOSPC"
+	echo "ERROR: dd not return ENOSPC"
+	sed "s/^/LOG: /" $LOG
 	SUCCESS=0
 fi
 
 # flush cache to OST(s) so avail numbers are correct
 sync; sleep 1 ; sync
 
-for OSC in /proc/fs/lustre/osc/OSC*MNT*; do
+for OSC in /proc/fs/lustre/osc/*-osc-*; do
 	AVAIL=`cat $OSC/kbytesavail`
-	GRANT=`cat $OSC/cur_grant_bytes`
-	[ $(($AVAIL - $GRANT / 1024)) -lt 400 ] && OSCFULL=full
+	GRANT=$((`cat $OSC/cur_grant_bytes` / 1024))
+	echo -n "$(basename $OSC) avl=$AVAIL grnt=$GRANT diff=$(($AVAIL - $GRANT))"
+	[ $(($AVAIL - $GRANT)) -lt 400 ] && OSCFULL=full && echo -n " FULL"
+	echo " "
 done
 
 if [ -z "$OSCFULL" ]; then
 	echo "no OSTs are close to full"
-	grep [0-9] /proc/fs/lustre/osc/OSC*MNT*/{kbytesavail,cur*}
+	grep "[0-9]" /proc/fs/lustre/osc/*-osc-*/{kbytesavail,cur*}
 	SUCCESS=0
 fi
 
 RECORDSOUT=`grep "records out" $LOG | cut -d + -f1`
-
 FILESIZE=`ls -l $OOS | awk '{ print $5 }'`
-if [ $RECORDSOUT -ne $(($FILESIZE / 1024)) ]; then
-        echo "ERROR: blocks written by dd not equal to the size of file"
-        SUCCESS=0
+if [ -z "$RECORDSOUT" ]; then
+	echo "ERROR: no blocks written by dd?"
+	sed "s/^/LOG: /" $LOG
+	SUCCESS=0
+elif [ "$RECORDSOUT" -ne $((FILESIZE / 1024)) ]; then
+	echo "ERROR: blocks written by dd not equal to the size of file"
+	SUCCESS=0
 fi
 
 #lctl debug_daemon stop
 
 rm -f $OOS
+sync; sleep 1; sync
+
+sync; sleep 3; sync
 
 if [ $SUCCESS -eq 1 ]; then
 	echo "Success!"
+	rm -f $LOG
 else
 	exit 1
 fi