LU-10729 tests: replay-dual/22d to wait

replay-dual/22d should have a similar procedure as 23d:
replay-dual/23d simulates a dropped reply for the executed
update, but previous tests can break this:
 - the update modifies remote llog
 - there can be another uptdate to that remote log
   (from the previous tests)
 - fail_loc (OBD_FAIL_UPDATE_OBJ_NET) is applied to the
   old update
 - the 23d's update gets stuck

so the test has to ensure there is no pending/in-flight
updates.

Test-Parameters: trivial testlist=replay-dual mdscount=2 mdtcount=4
Test-Parameters: testlist=replay-dual fstype=zfs mdscount=2 mdtcount=4
Signed-off-by: Xing Huang <hxing@ddn.com>
Change-Id: I2e3d3d4d1e5e33ffbb5c953edb21bcae884022c3
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/52343
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/tests/replay-dual.sh b/lustre/tests/replay-dual.sh
index 85836e6..2922aaf 100755 (executable)
--- a/lustre/tests/replay-dual.sh
+++ b/lustre/tests/replay-dual.sh
@@ -713,6 +713,12 @@ test_22d () {
        do_node $CLIENT1 $LFS mkdir -i 0 $MOUNT1/$tdir ||
                error "lfs mkdir -i 0 $MOUNT1/$tdir failed"
 
+       # let previous transactions to complete
+       # (distributed llog cancels, etc)
+       do_nodes $(comma_list $(mdts_nodes)) \
+               "$LCTL set_param -n osd*.*MDT*.force_sync=1"
+       sleep 2
+
        # OBD_FAIL_OUT_UPDATE_NET_REP    0x1701
        do_facet mds$MDTIDX lctl set_param fail_loc=0x1701
        do_node $CLIENT1 $LFS mkdir -i $MDTIDX $MOUNT1/$remote_dir &