2 files changed, 81 insertions, 4 deletions

diff --git a/qa/workunits/rbd/rbd_mirror.sh b/qa/workunits/rbd/rbd_mirror.sh
index f071c64e12e..71f8d7efae1 100755
--- a/qa/workunits/rbd/rbd_mirror.sh
+++ b/qa/workunits/rbd/rbd_mirror.sh
@@ -256,4 +256,51 @@ wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
 test_status_in_pool_dir ${CLUSTER1} ${POOL} ${image} 'up+replaying' 'master_position'
 compare_images ${POOL} ${image}
 
+testlog "TEST: client disconnect"
+image=laggy
+create_image ${CLUSTER2} ${POOL} ${image} 128 --journal-object-size 64K
+write_image ${CLUSTER2} ${POOL} ${image} 10
+
+testlog " - replay stopped after disconnect"
+wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
+wait_for_replay_complete ${CLUSTER1} ${CLUSTER2} ${POOL} ${image}
+test -n "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"
+disconnect_image ${CLUSTER2} ${POOL} ${image}
+test -z "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"
+wait_for_image_replay_stopped ${CLUSTER1} ${POOL} ${image}
+test_status_in_pool_dir ${CLUSTER1} ${POOL} ${image} 'up+error' 'disconnected'
+
+testlog " - replay started after resync requested"
+request_resync_image ${CLUSTER1} ${POOL} ${image}
+wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted'
+wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
+wait_for_replay_complete ${CLUSTER1} ${CLUSTER2} ${POOL} ${image}
+test -n "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"
+compare_images ${POOL} ${image}
+
+testlog " - disconnected after max_concurrent_object_sets reached"
+admin_daemon ${CLUSTER1} rbd mirror stop ${POOL}/${image}
+wait_for_image_replay_stopped ${CLUSTER1} ${POOL} ${image}
+test -n "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"
+set_image_meta ${CLUSTER2} ${POOL} ${image} \
+	       conf_rbd_journal_max_concurrent_object_sets 1
+write_image ${CLUSTER2} ${POOL} ${image} 20 16384
+write_image ${CLUSTER2} ${POOL} ${image} 20 16384
+test -z "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"
+set_image_meta ${CLUSTER2} ${POOL} ${image} \
+	       conf_rbd_journal_max_concurrent_object_sets 0
+
+testlog " - replay is still stopped (disconnected) after restart"
+admin_daemon ${CLUSTER1} rbd mirror start ${POOL}/${image}
+wait_for_image_replay_stopped ${CLUSTER1} ${POOL} ${image}
+test_status_in_pool_dir ${CLUSTER1} ${POOL} ${image} 'up+error' 'disconnected'
+
+testlog " - replay started after resync requested"
+request_resync_image ${CLUSTER1} ${POOL} ${image}
+wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted'
+wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
+wait_for_replay_complete ${CLUSTER1} ${CLUSTER2} ${POOL} ${image}
+test -n "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"
+compare_images ${POOL} ${image}
+
 echo OK
diff --git a/qa/workunits/rbd/rbd_mirror_helpers.sh b/qa/workunits/rbd/rbd_mirror_helpers.sh
index 8a362a16fbc..2b3e5807394 100755
--- a/qa/workunits/rbd/rbd_mirror_helpers.sh
+++ b/qa/workunits/rbd/rbd_mirror_helpers.sh
@@ -418,7 +418,7 @@ get_position()
     local status_log=${TEMPDIR}/${CLUSTER2}-${pool}-${image}.status
     rbd --cluster ${cluster} -p ${pool} journal status --image ${image} |
 	tee ${status_log} >&2
-    sed -nEe 's/^.*\[id='"${id_regexp}"',.*positions=\[\[([^]]*)\],.*$/\1/p' \
+    sed -nEe 's/^.*\[id='"${id_regexp}"',.*positions=\[\[([^]]*)\],.*state=connected.*$/\1/p' \
 	${status_log}
 }
 
@@ -489,12 +489,29 @@ test_status_in_pool_dir()
 
 create_image()
 {
+    local cluster=$1 ; shift
+    local pool=$1 ; shift
+    local image=$1 ; shift
+    local size=128
+
+    if [ -n "$1" ]; then
+	size=$1
+	shift
+    fi
+
+    rbd --cluster ${cluster} -p ${pool} create --size ${size} \
+	--image-feature layering,exclusive-lock,journaling $@ ${image}
+}
+
+set_image_meta()
+{
     local cluster=$1
     local pool=$2
     local image=$3
+    local key=$4
+    local val=$5
 
-    rbd --cluster ${cluster} -p ${pool} create --size 128 \
-	--image-feature layering,exclusive-lock,journaling ${image}
+    rbd --cluster ${cluster} -p ${pool} image-meta set ${image} $key $val
 }
 
 remove_image()
@@ -532,6 +549,16 @@ clone_image()
 	${clone_pool}/${clone_image} --image-feature layering,exclusive-lock,journaling
 }
 
+disconnect_image()
+{
+    local cluster=$1
+    local pool=$2
+    local image=$3
+
+    rbd --cluster ${cluster} -p ${pool} journal client disconnect \
+	--image ${image}
+}
+
 create_snapshot()
 {
     local cluster=$1
@@ -614,9 +641,12 @@ write_image()
     local pool=$2
     local image=$3
     local count=$4
+    local size=$5
+
+    test -n "${size}" || size=4096
 
     rbd --cluster ${cluster} -p ${pool} bench-write ${image} \
-	--io-size 4096 --io-threads 1 --io-total $((4096 * count)) \
+	--io-size ${size} --io-threads 1 --io-total $((size * count)) \
 	--io-pattern rand
 }