qa/workunits/rbd: use image id when probing for image presence

This fixes a race in resync tests leading to false negative results.

Fixes: http://tracker.ceph.com/issues/18048
Signed-off-by: Mykola Golub <mgolub@mirantis.com>

diff --git a/qa/workunits/rbd/rbd_mirror.sh b/qa/workunits/rbd/rbd_mirror.sh
index d1a963292a9112e847229e2f1c69135f4669f8ef..ee5755caac1789d5133e4bac8d7e12b4d9dd219b 100755 (executable)
--- a/qa/workunits/rbd/rbd_mirror.sh
+++ b/qa/workunits/rbd/rbd_mirror.sh
@@ -267,8 +267,8 @@ wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'present'
 wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
 
 testlog "TEST: simple image resync"
-request_resync_image ${CLUSTER1} ${POOL} ${image}
-wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted'
+request_resync_image ${CLUSTER1} ${POOL} ${image} image_id
+wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted' ${image_id}
 wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
 wait_for_status_in_pool_dir ${CLUSTER1} ${POOL} ${image} 'up+replaying' 'master_position'
 compare_images ${POOL} ${image}
@@ -276,9 +276,9 @@ compare_images ${POOL} ${image}
 testlog "TEST: image resync while replayer is stopped"
 admin_daemon ${CLUSTER1} rbd mirror stop ${POOL}/${image}
 wait_for_image_replay_stopped ${CLUSTER1} ${POOL} ${image}
-request_resync_image ${CLUSTER1} ${POOL} ${image}
+request_resync_image ${CLUSTER1} ${POOL} ${image} image_id
 admin_daemon ${CLUSTER1} rbd mirror start ${POOL}/${image}
-wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted'
+wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted' ${image_id}
 admin_daemon ${CLUSTER1} rbd mirror start ${POOL}/${image}
 wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
 wait_for_status_in_pool_dir ${CLUSTER1} ${POOL} ${image} 'up+replaying' 'master_position'
@@ -286,7 +286,7 @@ compare_images ${POOL} ${image}
 
 testlog "TEST: request image resync while daemon is offline"
 stop_mirror ${CLUSTER1}
-request_resync_image ${CLUSTER1} ${POOL} ${image}
+request_resync_image ${CLUSTER1} ${POOL} ${image} image_id
 start_mirror ${CLUSTER1}
 wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
 wait_for_status_in_pool_dir ${CLUSTER1} ${POOL} ${image} 'up+replaying' 'master_position'
@@ -307,8 +307,8 @@ wait_for_image_replay_stopped ${CLUSTER1} ${POOL} ${image}
 wait_for_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'
+request_resync_image ${CLUSTER1} ${POOL} ${image} image_id
+wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted' ${image_id}
 wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
 wait_for_replay_complete ${CLUSTER1} ${CLUSTER2} ${POOL} ${image}
 test -n "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"
@@ -332,8 +332,8 @@ wait_for_image_replay_stopped ${CLUSTER1} ${POOL} ${image}
 wait_for_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'
+request_resync_image ${CLUSTER1} ${POOL} ${image} image_id
+wait_for_image_present ${CLUSTER1} ${POOL} ${image} 'deleted' ${image_id}
 wait_for_image_replay_started ${CLUSTER1} ${POOL} ${image}
 wait_for_replay_complete ${CLUSTER1} ${CLUSTER2} ${POOL} ${image}
 test -n "$(get_mirror_position ${CLUSTER2} ${POOL} ${image})"

diff --git a/qa/workunits/rbd/rbd_mirror_helpers.sh b/qa/workunits/rbd/rbd_mirror_helpers.sh
index 44377f73e347d71a47cac75cc037263d93bc8ba5..6f51ca97dbc5273cdbaebf1e33cb06e8166a5cb8 100755 (executable)
--- a/qa/workunits/rbd/rbd_mirror_helpers.sh
+++ b/qa/workunits/rbd/rbd_mirror_helpers.sh
@@ -746,9 +746,13 @@ test_image_present()
     local pool=$2
     local image=$3
     local test_state=$4
+    local image_id=$5
     local current_state=deleted
+    local current_image_id
 
-    rbd --cluster=${cluster} -p ${pool} ls | grep "^${image}$" &&
+    current_image_id=$(get_image_id ${cluster} ${pool} ${image})
+    test -n "${current_image_id}" &&
+    test -z "${image_id}" -o "${image_id}" = "${current_image_id}" &&
     current_state=present
 
     test "${test_state}" = "${current_state}"
@@ -760,21 +764,41 @@ wait_for_image_present()
     local pool=$2
     local image=$3
     local state=$4
+    local image_id=$5
     local s
 
+    test -n "${image_id}" ||
+    image_id=$(get_image_id ${cluster} ${pool} ${image})
+
     # TODO: add a way to force rbd-mirror to update replayers
     for s in 0.1 1 2 4 8 8 8 8 8 8 8 8 16 16 32 32; do
        sleep ${s}
-       test_image_present "${cluster}" "${pool}" "${image}" "${state}" && return 0
+       test_image_present \
+            "${cluster}" "${pool}" "${image}" "${state}" "${image_id}" &&
+        return 0
     done
     return 1
 }
 
+get_image_id()
+{
+    local cluster=$1
+    local pool=$2
+    local image=$3
+
+    rbd --cluster=${cluster} -p ${pool} info ${image} |
+       sed -ne 's/^.*block_name_prefix: rbd_data\.//p'
+}
+
 request_resync_image()
 {
     local cluster=$1
     local pool=$2
     local image=$3
+    local image_id_var_name=$1
+
+    eval "${image_id_var_name}='$(get_image_id ${cluster} ${pool} ${image})'"
+    eval 'test -n "$'${image_id_var_name}'"'
 
     rbd --cluster=${cluster} -p ${pool} mirror image resync ${image}
 }