X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Foos.sh;h=98d2c183205fe7cbbda2313400746dee69bd08ba;hp=5894c4f53600baedeb8af04993fed8ec52bf8042;hb=fe2300b2fb91cc321412a6e53335848741ea30cd;hpb=cbd3a230274e1197d378725ee9e2aed6e35d2cb6 diff --git a/lustre/tests/oos.sh b/lustre/tests/oos.sh index 5894c4f..98d2c18 100755 --- a/lustre/tests/oos.sh +++ b/lustre/tests/oos.sh @@ -1,79 +1,100 @@ #!/bin/bash set -e -set -vx +#set -vx + +LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)} +. $LUSTRE/tests/test-framework.sh +. ${CONFIG:=$LUSTRE/tests/cfg/${NAME}.sh} export PATH=`dirname $0`/../utils:$PATH LFS=${LFS:-lfs} +LCTL=${LCTL:-lctl} MOUNT=${MOUNT:-$1} MOUNT=${MOUNT:-/mnt/lustre} OOS=$MOUNT/oosfile TMP=${TMP:-/tmp} -LOG=$TMP/ooslog +LOG=$TMP/$(basename $0 .sh).log SUCCESS=1 -rm -f $OOS +rm -f $OOS $LOG + +sync; sleep 1; sync # to ensure we get up-to-date statfs info -sleep 1 # to ensure we get up-to-date statfs info +#$LCTL set_param -n debug=-1 +#$LCTL set_param -n subsystem_debug=0x40a8 -#echo -1 > /proc/sys/portals/debug -#echo 0x40a8 > /proc/sys/portals/subsystem_debug -#lctl clear -#lctl debug_daemon start /r/tmp/debug 1024 +#$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 * $STRIPECOUNT))} +STRIPECOUNT=`$LCTL get_param -n lov.*.activeobd | head -n 1` +ORIGFREE=`$LCTL get_param -n llite.*.kbytesavail | head -n 1` +MAXFREE=${MAXFREE:-$((400000 * $STRIPECOUNT))} +echo STRIPECOUNT=$STRIPECOUNT ORIGFREE=$ORIGFREE MAXFREE=$MAXFREE if [ $ORIGFREE -gt $MAXFREE ]; then - echo "skipping out-of-space test on $OSC" - echo "reports ${ORIGFREE}kB free, more tham MAXFREE ${MAXFREE}kB" - echo "increase $MAXFREE (or reduce test fs size) to proceed" + skip "$0: ${ORIGFREE}kB free gt MAXFREE ${MAXFREE}kB, 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 +[ -f $LOG ] && error "log file wasn't removed?" + +echo BEFORE dd started +oos_full || true + # make sure we stripe over all OSTs to avoid OOS on only a subset of OSTs -$LFS setstripe $OOS 65536 0 $STRIPECOUNT -if dd if=/dev/zero of=$OOS count=$(($ORIGFREE + 100)) bs=1k 2> $LOG; then +$LFS setstripe $OOS -c $STRIPECOUNT +# 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> $LOG; then echo "ERROR: dd did not fail" SUCCESS=0 fi +[ ! -s "$LOG" ] && error "LOG file is empty!" + 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 - AVAIL=`cat $OSC/kbytesavail` - GRANT=`cat $OSC/cur_grant_bytes` - [ $(($AVAIL - $GRANT / 1024)) -lt 400 ] && OSCFULL=full -done -if [ -z "$OSCFULL" ]; then +echo AFTER dd +if ! oos_full; then echo "no OSTs are close to full" - grep [0-9] /proc/fs/lustre/osc/OSC*MNT*/{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 $((FILESIZE / 1024)) ]; then + echo "ERROR: blocks written by dd not equal to the size of file" + SUCCESS=0 fi -#lctl debug_daemon stop +#$LCTL debug_daemon stop +[ $SUCCESS != 0 ] && echo LOG file && sed "s/^/LOG: /" $LOG rm -f $OOS +sync; sleep 3; sync + +wait_delete_completed 300 + if [ $SUCCESS -eq 1 ]; then echo "Success!" + rm -f $LOG else exit 1 fi