[LU-15412] sanity test 807 stuck on multi-nodes setup when CLIENTS is empty

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0, Lustre 2.15.3
• Environment: Muli-node Lustre test environment.
• Severity: 3
• Assignee: Xinliang Liu
• Reporter: Xinliang Liu

Description

Sanity test 807 stuck on multi-nodes test environment when CLIENTS is empty.

[root@lustre-aio tests]# PTLDEBUG=-1  ./auster  -vr sanity --only 807
2...
3
4== sanity test 807: verify LSOM syncing tool ============= 02:19:11 (1641349151)
5mdd.lustre-MDT0000.changelog_mask=+hsm
6mdd.lustre-MDT0001.changelog_mask=+hsm
7Registered 2 changelog users: 'cl1 cl1'
8llite.lustre-fffff7ba8ec63000.xattr_cache=0
91+0 records in
101+0 records out
111048576 bytes (1.0 MB, 1.0 MiB) copied, 0.101602 s, 10.3 MB/s
12Test SOM for multi-client (0) writes
13No such rcmd module "(PATH=$PATH"
14pdsh@lustre-aio: Failed to register rcmd "(PATH=$PATH" for "/lib64/lustre/utils:/lib64/lustre/tests:/sbin:/usr/sbin; cd /lib64/lustre/tests; LUSTRE="/lib64/lustre"  mds1_FSTYPE=ldiskfs ost1_FSTYPE=ldiskfs VERBOSE=false FSTYPE=ldiskfs NETTYPE=tcp sh -c "sync ; sleep 5 ; sync")"
15Start to sync 3 records.
16lustre-MDT0000: clear the changelog for cl1 of all records
17lustre-MDT0000: Deregistered changelog user #1
18
19...stucking... 

But defining var “CLIENTS=$(hostname)” in cfg/local.sh, the test case passes.

 

sanity test 807 source code

27113 test_807() {
27114     [ -n "$FILESET" ] && skip "Not functional for FILESET set"
27115     [ $MDS1_VERSION -lt $(version_code 2.11.52) ] &&
27116         skip "Need MDS version at least 2.11.52"
27117
27118     # Registration step
27119     changelog_register || error "changelog_register failed"
27120     local cl_user="${CL_USERS[$SINGLEMDS]%% *}"
27121     changelog_users $SINGLEMDS | grep -q $cl_user ||
27122         error "User $cl_user not found in changelog_users"
27123
27124     local save="$TMP/$TESTSUITE-$TESTNAME.parameters"
27125     save_lustre_params client "llite.*.xattr_cache" > $save
27126     lctl set_param llite.*.xattr_cache=0
27127     stack_trap "restore_lustre_params < $save; rm -f $save" EXIT
27128
27129     rm -rf $DIR/$tdir || error "rm $tdir failed"
27130     mkdir_on_mdt0 $DIR/$tdir || error "mkdir $tdir failed"
27131     touch $DIR/$tdir/trunc || error "touch $tdir/trunc failed"
27132     $TRUNCATE $DIR/$tdir/trunc 1024 || error "truncate $tdir/trunc failed"
27133     $TRUNCATE $DIR/$tdir/trunc 1048576 ||
27134         error "truncate $tdir/trunc failed"
27135
27136     local bs=1048576
27137     dd if=/dev/zero of=$DIR/$tdir/single_dd bs=$bs count=1 conv=fsync ||
27138         error "write $tfile failed"
27139
27140     # multi-client wirtes
27141     local num=$(get_node_count ${CLIENTS//,/ })
27142     local offset=0
27143     local i=0
27144
27145     echo "Test SOM for multi-client ($num) writes"
27146     touch $DIR/$tfile || error "touch $tfile failed"
27147     $TRUNCATE $DIR/$tfile 0
27148     for client in ${CLIENTS//,/ }; do
27149         do_node $client $MULTIOP $DIR/$tfile Oz${offset}w${bs}c &
27150         local pids[$i]=$!
27151         i=$((i + 1))
27152         offset=$((offset + $bs))
27153     done
27154     for (( i=0; i < $num; i++ )); do
27155         wait ${pids[$i]}
27156     done
27157
27158     do_rpc_nodes "$CLIENTS" cancel_lru_locks osc
27159     do_nodes "$CLIENTS" "sync ; sleep 5 ; sync"
27160     $LSOM_SYNC -u $cl_user -m $FSNAME-MDT0000 $MOUNT
27161     check_lsom_data $DIR/$tdir/trunc
27162     check_lsom_data $DIR/$tdir/single_dd
27163     check_lsom_data $DIR/$tfile
27164
27165     rm -rf $DIR/$tdir
27166     # Deregistration step
27167     changelog_deregister || error "changelog_deregister failed"
27168 }
27169 run_test 807 "verify LSOM syncing tool"