test(negative): implement force drain node

Signed-off-by: Yang Chiu <[email protected]>

e2e/keywords/k8s.resource

@@ -30,3 +30,26 @@ Delete volume of ${workload_kind} ${workload_id} replica node
 
 Add deleted node back
     reboot_node_by_name    ${deleted_node}
+
+Force drain volume of ${workload_kind} ${workload_id} volume node
+    ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
+    ${volume_name} =    get_workload_volume_name    ${workload_name}
+    ${drained_node} =    get_volume_node    ${volume_name}
+    ${last_volume_node} =    get_volume_node    ${volume_name}
+    force_drain_node    ${drained_node}
+    wait_for_all_pods_evicted    ${drained_node}
+    Set Test Variable    ${drained_node}
+    Set Test Variable    ${last_volume_node}
+
+Force drain volume of ${workload_kind} ${workload_id} replica node
+    ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
+    ${volume_name} =    get_workload_volume_name    ${workload_name}
+    ${drained_node} =    get_replica_node    ${volume_name}
+    ${last_volume_node} =    get_volume_node    ${volume_name}
+    force_drain_node    ${drained_node}
+    wait_for_all_pods_evicted    ${drained_node}
+    Set Test Variable    ${drained_node}
+    Set Test Variable    ${last_volume_node}
+
+Uncordon the drained node
+    uncordon_node    ${drained_node}

e2e/keywords/workload.resource

@@ -46,7 +46,7 @@ Wait for volume of ${workload_kind} ${workload_id} healthy
     ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
     wait_for_workload_volume_healthy    ${workload_name}
 
-Wait until volume of ${workload_kind} ${workload_id} replica rebuidling started on ${replica_locality}
+Wait until volume of ${workload_kind} ${workload_id} replica rebuilding started on ${replica_locality}
     ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
     ${volume_name} =    get_workload_volume_name    ${workload_name}
     wait_for_replica_rebuilding_to_start_on_node    ${volume_name}    ${replica_locality}
@@ -66,6 +66,20 @@ Wait for volume of ${workload_kind} ${workload_id} attached and healthy
     ${volume_name} =    get_workload_volume_name    ${workload_name}
     wait_for_volume_healthy    ${volume_name}
 
+Wait for volume of ${workload_kind} ${workload_id} attached to the original node and degraded
+    ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
+    ${volume_name} =    get_workload_volume_name    ${workload_name}
+    wait_for_volume_degraded    ${volume_name}
+    ${volume_node} =    get_volume_node    ${volume_name}
+    Should Be Equal    ${last_volume_node}    ${volume_node}
+
+Wait for volume of ${workload_kind} ${workload_id} attached to another node and degraded
+    ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
+    ${volume_name} =    get_workload_volume_name    ${workload_name}
+    wait_for_volume_degraded    ${volume_name}
+    ${volume_node} =    get_volume_node    ${volume_name}
+    Should Not Be Equal    ${last_volume_node}    ${volume_node}
+
 Delete replica of ${workload_kind} ${workload_id} volume on ${replica_locality}
     ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
     ${volume_name} =    get_workload_volume_name    ${workload_name}

e2e/libs/k8s/k8s.py

@@ -1,12 +1,14 @@
 import time
 import subprocess
 import asyncio
+from kubernetes import client
 from workload.pod import create_pod
 from workload.pod import delete_pod
 from workload.pod import new_pod_manifest
 from workload.constant import IMAGE_UBUNTU
-
+from utility.utility import subprocess_exec_cmd
 from utility.utility import logging
+from utility.utility import get_retry_count_and_interval
 
 async def restart_kubelet(node_name, downtime_in_sec=10):
     manifest = new_pod_manifest(
@@ -24,5 +26,48 @@ async def restart_kubelet(node_name, downtime_in_sec=10):
 
 def delete_node(node_name):
     exec_cmd = ["kubectl", "delete", "node", node_name]
-    res = subprocess.check_output(exec_cmd)
-    logging(f"Executed command {exec_cmd} with result {res}")
+    res = subprocess_exec_cmd(exec_cmd)
+
+def drain_node(node_name):
+    exec_cmd = ["kubectl", "drain", node_name, "--ignore-daemonsets", "--delete-emptydir-data"]
+    res = subprocess_exec_cmd(exec_cmd)
+
+def force_drain_node(node_name):
+    exec_cmd = ["kubectl", "drain", node_name, "--force", "--ignore-daemonsets", "--delete-emptydir-data"]
+    res = subprocess_exec_cmd(exec_cmd)
+
+def cordon_node(node_name):
+    exec_cmd = ["kubectl", "cordon", node_name]
+    res = subprocess_exec_cmd(exec_cmd)
+
+def uncordon_node(node_name):
+    exec_cmd = ["kubectl", "uncordon", node_name]
+    res = subprocess_exec_cmd(exec_cmd)
+
+def get_all_pods_on_node(node_name):
+    api = client.CoreV1Api()
+    all_pods = api.list_namespaced_pod(namespace='longhorn-system', field_selector='spec.nodeName=' + node_name)
+    user_pods = [p for p in all_pods.items if (p.metadata.namespace != 'kube-system')]
+    return user_pods
+
+def wait_all_pods_evicted(node_name):
+    retry_count, retry_interval = get_retry_count_and_interval()
+    for i in range(retry_count):
+        pods = get_all_pods_on_node(node_name)
+        logging(f"Waiting for pods evicted from {node_name} ... ({i})")
+        evicted = True
+        for pod in pods:
+            # check non DaemonSet Pods are evicted or terminating (deletionTimestamp != None)
+            pod_type = pod.metadata.owner_references[0].kind
+            pod_delete_timestamp = pod.metadata.deletion_timestamp
+
+            if (pod_type != 'DaemonSet' and pod_type != 'BackingImageManager') and pod_delete_timestamp == None:
+                evicted = False
+                break
+
+        if evicted:
+             break
+
+        time.sleep(retry_interval)
+
+    assert evicted, 'failed to evict pods'

e2e/libs/keywords/k8s_keywords.py

@@ -2,6 +2,9 @@
 from robot.libraries.BuiltIn import BuiltIn
 from k8s.k8s import restart_kubelet
 from k8s.k8s import delete_node
+from k8s.k8s import drain_node, force_drain_node
+from k8s.k8s import cordon_node, uncordon_node
+from k8s.k8s import wait_all_pods_evicted
 from utility.utility import logging
 
 
@@ -37,3 +40,15 @@ def delete_replica_node(self, volume_name):
         replica_node = volume_keywords.get_replica_node(volume_name)
         delete_node(replica_node)
         return replica_node
+
+    def drain_node(self, node_name):
+        drain_node(node_name)
+
+    def force_drain_node(self, node_name):
+        force_drain_node(node_name)
+
+    def uncordon_node(self, node_name):
+        uncordon_node(node_name)
+
+    def wait_for_all_pods_evicted(self, node_name):
+        wait_all_pods_evicted(node_name)

e2e/libs/node/node.py

@@ -49,33 +49,6 @@ def reset_disks(self, node_name):
                 logging(f"Try to remove disk {disk_name} from node {node_name}")
         self.update_disks(node_name, disks)
 
-    def get_all_pods_on_node(self, node_name):
-        api = client.CoreV1Api()
-        all_pods = api.list_namespaced_pod(namespace='longhorn-system', field_selector='spec.nodeName=' + node_name)
-        user_pods = [p for p in all_pods.items if (p.metadata.namespace != 'kube-system')]
-        return user_pods
-
-    def wait_all_pods_evicted(self, node_name):
-        retry_count, retry_interval = get_retry_count_and_interval()
-        for _ in range(retry_count):
-            pods = self.get_all_pods_on_node(node_name)
-            evicted = True
-            for pod in pods:
-                # check non DaemonSet Pods are evicted or terminating (deletionTimestamp != None)
-                pod_type = pod.metadata.owner_references[0].kind
-                pod_delete_timestamp = pod.metadata.deletion_timestamp
-
-                if pod_type != 'DaemonSet' and pod_delete_timestamp == None:
-                    evicted = False
-                    break
-
-            if evicted:
-                break
-
-            time.sleep(retry_interval)
-
-        assert evicted, 'failed to evict pods'
-
     def is_accessing_node_by_index(self, node):
         p = re.compile('node (\d)')
         if m := p.match(node):

e2e/libs/utility/utility.py

@@ -7,6 +7,7 @@
 import signal
 from robot.api import logger
 from robot.libraries.BuiltIn import BuiltIn
+import subprocess
 
 from longhorn import from_env
 
@@ -79,6 +80,12 @@ def get_backupstore():
     return os.environ.get('LONGHORN_BACKUPSTORE', "")
 
 
+def subprocess_exec_cmd(cmd):
+    res = subprocess.check_output(cmd)
+    logging(f"Executed command {cmd} with result {res}")
+    return res
+
+
 def wait_for_cluster_ready():
     core_api = client.CoreV1Api()
     retry_count, retry_interval = get_retry_count_and_interval()

e2e/tests/node_delete.robot

@@ -26,7 +26,7 @@ Delete Volume Node While Replica Rebuilding
 
     FOR    ${i}    IN RANGE    ${LOOP_COUNT}
         When Delete replica of deployment 0 volume on volume node
-        And Wait until volume of deployment 0 replica rebuidling started on volume node
+        And Wait until volume of deployment 0 replica rebuilding started on volume node
         And Delete volume of deployment 0 volume node
 
         Then Wait for volume of deployment 0 attached and unknown
@@ -44,7 +44,7 @@ Delete Replica Node While Replica Rebuilding
 
     FOR    ${i}    IN RANGE    ${LOOP_COUNT}
         When Delete replica of deployment 0 volume on replica node
-        And Wait until volume of deployment 0 replica rebuidling started on replica node
+        And Wait until volume of deployment 0 replica rebuilding started on replica node
         And Delete volume of deployment 0 replica node
 
         Then Wait for volume of deployment 0 attached and degraded

e2e/tests/node_drain.robot

@@ -0,0 +1,53 @@
+*** Settings ***
+Documentation    Negative Test Cases
+
+Resource    ../keywords/common.resource
+Resource    ../keywords/persistentvolumeclaim.resource
+Resource    ../keywords/k8s.resource
+Resource    ../keywords/deployment.resource
+Resource    ../keywords/workload.resource
+
+Test Setup    Set test environment
+Test Teardown    Cleanup test resources
+
+*** Variables ***
+${LOOP_COUNT}    1
+${RETRY_COUNT}    300
+${RETRY_INTERVAL}    1
+
+*** Test Cases ***
+Force Drain Volume Node While Replica Rebuilding
+    Given Create persistentvolumeclaim 0 using RWO volume
+    And Create deployment 0 with persistentvolumeclaim 0
+    And Wait for volume of deployment 0 healthy
+    And Write 2048 MB data to file data.txt in deployment 0
+
+    FOR    ${i}    IN RANGE    ${LOOP_COUNT}
+        When Delete replica of deployment 0 volume on volume node
+        And Wait until volume of deployment 0 replica rebuilding started on volume node
+        And Force drain volume of deployment 0 volume node
+
+        Then Wait for volume of deployment 0 attached to another node and degraded
+        And Uncordon the drained node
+        And Wait for volume of deployment 0 attached and healthy
+        And Wait for deployment 0 pods stable
+        And Check deployment 0 data in file data.txt is intact
+    END
+
+Force Drain Replica Node While Replica Rebuilding
+    Given Create persistentvolumeclaim 0 using RWO volume
+    And Create deployment 0 with persistentvolumeclaim 0
+    And Wait for volume of deployment 0 healthy
+    And Write 2048 MB data to file data.txt in deployment 0
+
+    FOR    ${i}    IN RANGE    ${LOOP_COUNT}
+        When Delete replica of deployment 0 volume on replica node
+        And Wait until volume of deployment 0 replica rebuilding started on replica node
+        And Force drain volume of deployment 0 replica node
+
+        Then Wait for volume of deployment 0 attached to the original node and degraded
+        And Uncordon the drained node
+        And Wait for volume of deployment 0 attached and healthy
+        And Wait for deployment 0 pods stable
+        And Check deployment 0 data in file data.txt is intact
+    END

e2e/tests/replica_rebuilding.robot

@@ -3,7 +3,6 @@ Documentation    Negative Test Cases
 
 Resource    ../keywords/common.resource
 Resource    ../keywords/host.resource
-Resource    ../keywords/persistentvolumeclaim.resource
 Resource    ../keywords/volume.resource
 
 Test Setup    Set test environment