return 0
}
+function _scrub_snaps_multi() {
+ local dir=$1
+ local poolname=test
+ local OBJS=15
+ local OSDS=2
+ local which=$2
+
+ TESTDATA="testdata.$$"
+
+ run_mon $dir a --osd_pool_default_size=$OSDS || return 1
+ run_mgr $dir x || return 1
+ for osd in $(seq 0 $(expr $OSDS - 1))
+ do
+ run_osd $dir $osd || return 1
+ done
+
+ # Create a pool with a single pg
+ create_pool $poolname 1 1
+ wait_for_clean || return 1
+ poolid=$(ceph osd dump | grep "^pool.*[']test[']" | awk '{ print $2 }')
+
+ dd if=/dev/urandom of=$TESTDATA bs=1032 count=1
+ for i in `seq 1 $OBJS`
+ do
+ rados -p $poolname put obj${i} $TESTDATA
+ done
+
+ local primary=$(get_primary $poolname obj1)
+ local replica=$(get_not_primary $poolname obj1)
+
+ eval create_scenario $dir $poolname $TESTDATA \$$which
+
+ rm -f $TESTDATA
+
+ for osd in $(seq 0 $(expr $OSDS - 1))
+ do
+ run_osd $dir $osd || return 1
+ done
+
+ local pgid="${poolid}.0"
+ if ! pg_scrub "$pgid" ; then
+ cat $dir/osd.0.log
+ return 1
+ fi
+
+ test "$(grep "_scan_snaps start" $dir/osd.${primary}.log | wc -l)" -gt "3" || return 1
+ test "$(grep "_scan_snaps start" $dir/osd.${replica}.log | wc -l)" -gt "3" || return 1
+
+ rados list-inconsistent-pg $poolname > $dir/json || return 1
+ # Check pg count
+ test $(jq '. | length' $dir/json) = "1" || return 1
+ # Check pgid
+ test $(jq -r '.[0]' $dir/json) = $pgid || return 1
+
+ rados list-inconsistent-obj $pgid --format=json-pretty
+
+ rados list-inconsistent-snapset $pgid > $dir/json || return 1
+
+ # Since all of the snapshots on the primary is consistent there are no errors here
+ if [ $which = "replica" ];
+ then
+ scruberrors="20"
+ jq "$jqfilter" << EOF | python -c "$sortkeys" > $dir/checkcsjson
+{
+ "epoch": 23,
+ "inconsistents": []
+}
+EOF
+
+else
+ scruberrors="30"
+ jq "$jqfilter" << EOF | python -c "$sortkeys" > $dir/checkcsjson
+{
+ "epoch": 23,
+ "inconsistents": [
+ {
+ "name": "obj10",
+ "nspace": "",
+ "locator": "",
+ "snap": 1,
+ "errors": [
+ "size_mismatch"
+ ]
+ },
+ {
+ "name": "obj11",
+ "nspace": "",
+ "locator": "",
+ "snap": 1,
+ "errors": [
+ "headless"
+ ]
+ },
+ {
+ "name": "obj14",
+ "nspace": "",
+ "locator": "",
+ "snap": 1,
+ "errors": [
+ "size_mismatch"
+ ]
+ },
+ {
+ "name": "obj6",
+ "nspace": "",
+ "locator": "",
+ "snap": 1,
+ "errors": [
+ "headless"
+ ]
+ },
+ {
+ "name": "obj7",
+ "nspace": "",
+ "locator": "",
+ "snap": 1,
+ "errors": [
+ "headless"
+ ]
+ },
+ {
+ "name": "obj9",
+ "nspace": "",
+ "locator": "",
+ "snap": 1,
+ "errors": [
+ "size_mismatch"
+ ]
+ },
+ {
+ "name": "obj5",
+ "nspace": "",
+ "locator": "",
+ "snap": 7,
+ "errors": [
+ "info_missing",
+ "headless"
+ ]
+ },
+ {
+ "name": "obj10",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 1,
+ "snaps": [
+ 1
+ ]
+ },
+ "clones": [
+ {
+ "snap": 1,
+ "size": 1032,
+ "overlap": "????",
+ "snaps": [
+ 1
+ ]
+ }
+ ]
+ },
+ "errors": []
+ },
+ {
+ "name": "obj11",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 1,
+ "snaps": [
+ 1
+ ]
+ },
+ "clones": []
+ },
+ "errors": [
+ "extra_clones"
+ ],
+ "extra clones": [
+ 1
+ ]
+ },
+ {
+ "name": "obj14",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 1,
+ "snaps": [
+ 1
+ ]
+ },
+ "clones": [
+ {
+ "snap": 1,
+ "size": 1033,
+ "overlap": "[]",
+ "snaps": [
+ 1
+ ]
+ }
+ ]
+ },
+ "errors": []
+ },
+ {
+ "name": "obj5",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 6,
+ "snaps": [
+ 6,
+ 5,
+ 4,
+ 3,
+ 2,
+ 1
+ ]
+ },
+ "clones": [
+ {
+ "snap": 1,
+ "size": 1032,
+ "overlap": "[]",
+ "snaps": [
+ 1
+ ]
+ },
+ {
+ "snap": 2,
+ "size": 256,
+ "overlap": "[]",
+ "snaps": [
+ 2
+ ]
+ },
+ {
+ "snap": 4,
+ "size": 512,
+ "overlap": "[]",
+ "snaps": [
+ 4,
+ 3
+ ]
+ },
+ {
+ "snap": 6,
+ "size": 1024,
+ "overlap": "[]",
+ "snaps": [
+ 6,
+ 5
+ ]
+ }
+ ]
+ },
+ "errors": [
+ "extra_clones"
+ ],
+ "extra clones": [
+ 7
+ ]
+ },
+ {
+ "name": "obj6",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 1,
+ "snaps": [
+ 1
+ ]
+ },
+ "clones": []
+ },
+ "errors": [
+ "extra_clones"
+ ],
+ "extra clones": [
+ 1
+ ]
+ },
+ {
+ "name": "obj7",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 0,
+ "snaps": []
+ },
+ "clones": []
+ },
+ "errors": [
+ "extra_clones"
+ ],
+ "extra clones": [
+ 1
+ ]
+ },
+ {
+ "name": "obj8",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 0,
+ "snaps": [
+ 1
+ ]
+ },
+ "clones": [
+ {
+ "snap": 1,
+ "size": 1032,
+ "overlap": "[]",
+ "snaps": [
+ 1
+ ]
+ }
+ ]
+ },
+ "errors": [
+ "snapset_error"
+ ]
+ },
+ {
+ "name": "obj9",
+ "nspace": "",
+ "locator": "",
+ "snap": "head",
+ "snapset": {
+ "snap_context": {
+ "seq": 1,
+ "snaps": [
+ 1
+ ]
+ },
+ "clones": [
+ {
+ "snap": 1,
+ "size": "????",
+ "overlap": "[]",
+ "snaps": [
+ 1
+ ]
+ }
+ ]
+ },
+ "errors": []
+ }
+ ]
+}
+EOF
+fi
+
+ jq "$jqfilter" $dir/json | python -c "$sortkeys" > $dir/csjson
+ diff ${DIFFCOLOPTS} $dir/checkcsjson $dir/csjson || test $getjson = "yes" || return 1
+ if test $getjson = "yes"
+ then
+ jq '.' $dir/json > save1.json
+ fi
+
+ if test "$LOCALRUN" = "yes" && which jsonschema > /dev/null;
+ then
+ jsonschema -i $dir/json $CEPH_ROOT/doc/rados/command/list-inconsistent-snap.json || return 1
+ fi
+
+ pidfiles=$(find $dir 2>/dev/null | grep 'osd[^/]*\.pid')
+ pids=""
+ for pidfile in ${pidfiles}
+ do
+ pids+="$(cat $pidfile) "
+ done
+
+ # When removing snapshots with a corrupt replica, it crashes.
+ # See http://tracker.ceph.com/issues/23875
+ if [ $which = "primary" ];
+ then
+ for i in `seq 1 7`
+ do
+ rados -p $poolname rmsnap snap$i
+ done
+ fi
+
+ ERRORS=0
+
+ for pid in $pids
+ do
+ if ! kill -0 $pid
+ then
+ echo "OSD Crash occurred"
+ ERRORS=$(expr $ERRORS + 1)
+ fi
+ done
+
+ kill_daemons $dir || return 1
+
+ declare -a err_strings
+ err_strings[0]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 shard [0-1] missing .*:::obj4:7"
+ err_strings[1]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 shard [0-1]: soid .*:::obj3:head size 3840 != size 768 from auth oi"
+ err_strings[2]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 shard [0-1] missing .*:::obj5:1"
+ err_strings[3]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 shard [0-1] missing .*:::obj5:2"
+ err_strings[4]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 shard [0-1]: soid .*:::obj5:4 size 4608 != size 512 from auth oi"
+ err_strings[5]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 soid .*:::obj5:7: failed to pick suitable object info"
+ err_strings[6]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 shard [0-1] missing .*:::obj1:head"
+ err_strings[7]="log_channel[(]cluster[)] log [[]ERR[]] : [0-9]*[.]0 scrub ${scruberrors} errors"
+
+ for err_string in "${err_strings[@]}"
+ do
+ if ! grep "$err_string" $dir/osd.${primary}.log > /dev/null;
+ then
+ echo "Missing log message '$err_string'"
+ ERRORS=$(expr $ERRORS + 1)
+ fi
+ done
+
+ if [ $ERRORS != "0" ];
+ then
+ echo "TEST FAILED WITH $ERRORS ERRORS"
+ return 1
+ fi
+
+ echo "TEST PASSED"
+ return 0
+}
+
+function TEST_scrub_snaps_replica() {
+ local dir=$1
+ ORIG_ARGS=$CEPH_ARGS
+ CEPH_ARGS+=" --osd_scrub_chunk_min=3 --osd_scrub_chunk_max=3"
+ _scrub_snaps_multi $dir replica
+ err=$?
+ CEPH_ARGS=$ORIG_ARGS
+ return $err
+}
+
+function TEST_scrub_snaps_primary() {
+ local dir=$1
+ ORIG_ARGS=$CEPH_ARGS
+ CEPH_ARGS+=" --osd_scrub_chunk_min=3 --osd_scrub_chunk_max=3"
+ _scrub_snaps_multi $dir primary
+ err=$?
+ CEPH_ARGS=$ORIG_ARGS
+ return $err
+}
+
main osd-scrub-snaps "$@"
# Local Variables: