Details
-
Bug
-
Resolution: Cannot Reproduce
-
Minor
-
None
-
Lustre 2.8.0
-
None
-
Eagle cluster with Lustre 2.7.64
-
3
-
9223372036854775807
Description
Running the sanity-krb5 test suite on Lustre systems with a separate MGS and MDS on the separate nodes, test 90 fails with
'dbench exit with error'
We set SLOW=yes for this set of tests. For sanity-krb5 test 90, that means that we call 'lfs flushctx' on the mount point 6 times more than we not running with the LSOW flag. From test 90:
if [ "$SLOW" = "no" ]; then
total=10
else
total=60
fi
restore_to_default_flavor
set_rule $FSNAME any any krb5p
wait_flavor all2all krb5p
start_dbench
for ((n=0;n<$total;n++)); do
sleep 2
check_dbench
echo "flush ctx ($n/$total) ..."
$LFS flushctx $MOUNT || error "can't flush context on $MOUNT"
done
From the test logs, debench had finished:
... flush ctx (13/60) ... flush ctx (14/60) ... flush ctx (15/60) ... flush ctx (16/60) ... flush ctx (17/60) ... flush ctx (18/60) ... dbench 19216 already finished sanity-krb5 test_90: @@@@@@ FAIL: dbench exit with error
This test fails infrequently.
Logs are at:
https://testing.hpdd.intel.com/test_sets/a246e266-b987-11e5-a748-5254006e85c2