fix: attach volumes to eliminate duplicated replicas

ref: longhorn/longhorn 8573

Signed-off-by: Jack Lin <jack.lin@suse.com>

build-image.sh

@@ -20,9 +20,9 @@ esac
 echo "Building for ${ARCH}"
 # update base image to get latest changes
 BASE_IMAGE=`grep FROM package/Dockerfile  | awk '{print $2}'`
-docker pull ${BASE_IMAGE}
+sudo docker pull ${BASE_IMAGE}
 
-docker build --build-arg TARGETPLATFORM="linux/${ARCH}" -t ${image} -f manager/integration/Dockerfile .
+sudo docker build --build-arg TARGETPLATFORM="linux/${ARCH}" -t ${image} -f manager/integration/Dockerfile .
 mkdir -p bin
 echo ${image} > bin/latest_image
 echo Built image ${image}

manager/integration/tests/common.py

@@ -6494,3 +6494,14 @@ def check_backing_image_eviction_failed(name): # NOQA
         time.sleep(RETRY_INTERVAL)
 
     assert check
+
+
+def wait_for_replica_count(client, volume_name, replica_count):
+    for i in range(RETRY_COUNTS):
+        volume = client.by_id_volume(volume_name)
+        if len(volume.replicas) == replica_count:
+            break
+        time.sleep(RETRY_INTERVAL)
+
+    volume = client.by_id_volume(volume_name)
+    assert volume.replicaCount == replica_count

manager/integration/tests/test_ha.py

@@ -83,6 +83,7 @@
 from common import wait_scheduling_failure
 from common import set_tags_for_node_and_its_disks
 from common import wait_for_tainted_node_engine_image_undeployed
+from common import wait_for_replica_count
 
 from backupstore import set_random_backupstore # NOQA
 from backupstore import backupstore_cleanup
@@ -2877,11 +2878,13 @@ def test_engine_image_not_fully_deployed_perform_auto_upgrade_engine(client, cor
     tainted node and not fully deployed engine.
 
     1. Create 2 volumes vol-1 and vol-2 with 2 replicas
-    2. Deploy a new engine image, new-ei
-    3. Upgrade vol-1 and vol-2 to the new-ei
-    4. Attach vol-2 to current-node
-    5. Set `Concurrent Automatic Engine Upgrade Per Node Limit` setting to 3
-    6. In a 2-min retry, verify that Longhorn upgrades the engine image of
+    2. Attach both volumes to make sure they are healthy and have 2 replicas
+    4. Detach both volumes
+    5. Deploy a new engine image, new-ei
+    6. Upgrade vol-1 and vol-2 to the new-ei
+    7. Attach vol-2 to current-node
+    8. Set `Concurrent Automatic Engine Upgrade Per Node Limit` setting to 3
+    9. In a 2-min retry, verify that Longhorn upgrades the engine image of
        vol-1 and vol-2.
     """
     tainted_node_id = \
@@ -2895,6 +2898,18 @@ def test_engine_image_not_fully_deployed_perform_auto_upgrade_engine(client, cor
                                       num_of_replicas=2,
                                       size=str(3 * Gi))
 
+    volume1.attach(hostId=get_self_host_id())
+    volume2.attach(hostId=get_self_host_id())
+    volume1 = wait_for_volume_healthy(client, volume1.name)
+    volume2 = wait_for_volume_healthy(client, volume2.name)
+    wait_for_replica_count(client, volume1.name, 2)
+    wait_for_replica_count(client, volume2.name, 2)
+
+    volume1.detach()
+    volume2.detach()
+    volume1 = wait_for_volume_detached(client, volume1.name)
+    volume2 = wait_for_volume_detached(client, volume2.name)
+
     default_img = common.get_default_engine_image(client)
     # engine reference =
     # (1 volume + 1 engine + number of replicas) * volume count