Adding node_label_selector for pod scenarios (#888)

* Adding node_label_selector for pod scenarios

Signed-off-by: Sahil Shah <sahshah@redhat.com>

* using kubernetes function, adding node_name and removing extra config

Signed-off-by: Sahil Shah <sahshah@redhat.com>

* adding CI test for custom pod scenario

Signed-off-by: Sahil Shah <sahshah@redhat.com>

* fixing comment

* adding test to workflow

* adding list parsing logic for krkn hub

* parsing not needed, as input is always []

---------

Signed-off-by: Sahil Shah <sahshah@redhat.com>

.github/workflows/tests.yml

@@ -73,6 +73,7 @@ jobs:
             echo "test_app_outages" >> ./CI/tests/functional_tests
             echo "test_container"      >> ./CI/tests/functional_tests
             echo "test_pod" >> ./CI/tests/functional_tests
+            echo "test_customapp_pod" >> ./CI/tests/functional_tests
             echo "test_namespace"      >> ./CI/tests/functional_tests
             echo "test_net_chaos"      >> ./CI/tests/functional_tests
             echo "test_time"           >> ./CI/tests/functional_tests
@@ -108,6 +109,7 @@ jobs:
           echo "test_app_outages" >> ./CI/tests/functional_tests
           echo "test_container"      >> ./CI/tests/functional_tests
           echo "test_pod" >> ./CI/tests/functional_tests
+          echo "test_customapp_pod" >> ./CI/tests/functional_tests
           echo "test_namespace"      >> ./CI/tests/functional_tests
           echo "test_net_chaos"      >> ./CI/tests/functional_tests
           echo "test_time"           >> ./CI/tests/functional_tests

CI/tests/test_customapp_pod.sh

@@ -0,0 +1,18 @@
+set -xeEo pipefail
+
+source CI/tests/common.sh
+
+trap error ERR
+trap finish EXIT
+
+function functional_test_customapp_pod_node_selector {
+  export scenario_type="pod_disruption_scenarios"
+  export scenario_file="scenarios/openshift/customapp_pod.yaml"
+  export post_config=""
+  envsubst < CI/config/common_test_config.yaml > CI/config/customapp_pod_config.yaml
+
+  python3 -m coverage run -a run_kraken.py -c CI/config/customapp_pod_config.yaml
+  echo "Pod disruption with node_label_selector test: Success"
+}
+
+functional_test_customapp_pod_node_selector

krkn/scenario_plugins/pod_disruption/models/models.py

@@ -11,6 +11,8 @@ class InputParams:
             self.label_selector = config["label_selector"] if "label_selector" in config else ""
             self.namespace_pattern = config["namespace_pattern"] if "namespace_pattern" in config else ""
             self.name_pattern = config["name_pattern"] if "name_pattern" in config else ""
+            self.node_label_selector = config["node_label_selector"] if "node_label_selector" in config else ""
+            self.node_names = config["node_names"] if "node_names" in config else []
 
     namespace_pattern: str
     krkn_pod_recovery_time: int
@@ -18,4 +20,6 @@ class InputParams:
     duration: int
     kill: int
     label_selector: str
-    name_pattern: str
+    name_pattern: str
+    node_label_selector: str
+    node_names: list

krkn/scenario_plugins/pod_disruption/pod_disruption_scenario_plugin.py

@@ -100,18 +100,86 @@ class PodDisruptionScenarioPlugin(AbstractScenarioPlugin):
             raise Exception(
                 f"impossible to determine monitor parameters, check {kill_scenario} configuration"
             )
+    
+    def _select_pods_with_field_selector(self, name_pattern, label_selector, namespace, kubecli: KrknKubernetes, field_selector: str, node_name: str = None):
+        """Helper function to select pods using either label_selector or name_pattern with field_selector, optionally filtered by node"""
+        # Combine field selectors if node targeting is specified
+        if node_name:
+            node_field_selector = f"spec.nodeName={node_name}"
+            if field_selector:
+                combined_field_selector = f"{field_selector},{node_field_selector}"
+            else:
+                combined_field_selector = node_field_selector
+        else:
+            combined_field_selector = field_selector
         
+        if label_selector:
+            return kubecli.select_pods_by_namespace_pattern_and_label(
+                label_selector=label_selector, 
+                namespace_pattern=namespace, 
+                field_selector=combined_field_selector
+            )
+        else:  # name_pattern
+            return kubecli.select_pods_by_name_pattern_and_namespace_pattern(
+                pod_name_pattern=name_pattern, 
+                namespace_pattern=namespace, 
+                field_selector=combined_field_selector
+            )
 
-    def get_pods(self, name_pattern, label_selector,namespace, kubecli: KrknKubernetes, field_selector: str =None): 
+    def get_pods(self, name_pattern, label_selector, namespace, kubecli: KrknKubernetes, field_selector: str = None, node_label_selector: str = None, node_names: list = None, quiet: bool = False): 
         if label_selector and name_pattern: 
             logging.error('Only, one of name pattern or label pattern can be specified')
-        elif label_selector:
-            pods = kubecli.select_pods_by_namespace_pattern_and_label(label_selector=label_selector,namespace_pattern=namespace, field_selector=field_selector)
-        elif name_pattern:
-            pods = kubecli.select_pods_by_name_pattern_and_namespace_pattern(pod_name_pattern=name_pattern, namespace_pattern=namespace, field_selector=field_selector)
-        else: 
+            return []
+        
+        if not label_selector and not name_pattern:
             logging.error('Name pattern or label pattern must be specified ')
-        return pods
+            return []
+        
+        # If specific node names are provided, make multiple calls with field selector
+        if node_names:
+            if not quiet:
+                logging.info(f"Targeting pods on {len(node_names)} specific nodes")
+            all_pods = []
+            for node_name in node_names:
+                pods = self._select_pods_with_field_selector(
+                    name_pattern, label_selector, namespace, kubecli, field_selector, node_name
+                )
+                
+                if pods:
+                    all_pods.extend(pods)
+            
+            if not quiet:
+                logging.info(f"Found {len(all_pods)} target pods across {len(node_names)} nodes")
+            return all_pods
+        
+        #  Node label selector approach - use field selectors
+        if node_label_selector:
+            # Get nodes matching the label selector first
+            nodes_with_label = kubecli.list_nodes(label_selector=node_label_selector)
+            if not nodes_with_label:
+                logging.info(f"No nodes found with label selector: {node_label_selector}")
+                return []
+            
+            if not quiet:
+                logging.info(f"Targeting pods on {len(nodes_with_label)} nodes with label: {node_label_selector}")
+            # Use field selector for each node
+            all_pods = []
+            for node_name in nodes_with_label:
+                pods = self._select_pods_with_field_selector(
+                    name_pattern, label_selector, namespace, kubecli, field_selector, node_name
+                )
+                
+                if pods:
+                    all_pods.extend(pods)
+            
+            if not quiet:
+                logging.info(f"Found {len(all_pods)} target pods across {len(nodes_with_label)} nodes")
+            return all_pods
+        
+        # Standard pod selection (no node targeting)
+        return self._select_pods_with_field_selector(
+            name_pattern, label_selector, namespace, kubecli, field_selector
+        )
     
     def killing_pods(self, config: InputParams, kubecli: KrknKubernetes):
         # region Select target pods
@@ -120,7 +188,7 @@ class PodDisruptionScenarioPlugin(AbstractScenarioPlugin):
         if not namespace: 
             logging.error('Namespace pattern must be specified')
 
-        pods = self.get_pods(config.name_pattern,config.label_selector,config.namespace_pattern, kubecli, field_selector="status.phase=Running")
+        pods = self.get_pods(config.name_pattern,config.label_selector,config.namespace_pattern, kubecli, field_selector="status.phase=Running", node_label_selector=config.node_label_selector, node_names=config.node_names)
         pods_count = len(pods)
         if len(pods) < config.kill:
             logging.error("Not enough pods match the criteria, expected {} but found only {} pods".format(
@@ -129,23 +197,22 @@ class PodDisruptionScenarioPlugin(AbstractScenarioPlugin):
         
         random.shuffle(pods)
         for i in range(config.kill):
-
             pod = pods[i]
             logging.info(pod)
             logging.info(f'Deleting pod {pod[0]}')
             kubecli.delete_pod(pod[0], pod[1])
         
-        self.wait_for_pods(config.label_selector,config.name_pattern,config.namespace_pattern, pods_count, config.duration, config.timeout, kubecli)
+        self.wait_for_pods(config.label_selector,config.name_pattern,config.namespace_pattern, pods_count, config.duration, config.timeout, kubecli, config.node_label_selector, config.node_names)
         return 0
 
     def wait_for_pods(
-        self, label_selector, pod_name, namespace, pod_count, duration, wait_timeout, kubecli: KrknKubernetes
+        self, label_selector, pod_name, namespace, pod_count, duration, wait_timeout, kubecli: KrknKubernetes, node_label_selector, node_names
     ):
         timeout = False
         start_time = datetime.now()
 
         while not timeout:
-            pods = self.get_pods(name_pattern=pod_name, label_selector=label_selector,namespace=namespace, field_selector="status.phase=Running", kubecli=kubecli)
+            pods = self.get_pods(name_pattern=pod_name, label_selector=label_selector,namespace=namespace, field_selector="status.phase=Running", kubecli=kubecli, node_label_selector=node_label_selector, node_names=node_names, quiet=True)
             if pod_count == len(pods):
                 return
                

scenarios/openshift/customapp_pod.yaml

@@ -1,6 +1,15 @@
 # yaml-language-server: $schema=../plugin.schema.json
 - id: kill-pods
   config:
-    namespace_pattern: ^acme-air$
+    namespace_pattern: "kube-system"
     name_pattern: .*
-    krkn_pod_recovery_time: 120
+    krkn_pod_recovery_time: 60
+    kill: 1     # num of pods to kill
+    #Not needed by default, but can be used if you want to target pods on specific nodes
+    # Option 1: Target pods on nodes with specific labels [master/worker nodes]
+    node_label_selector: node-role.kubernetes.io/control-plane=      # Target control-plane nodes (works on both k8s and openshift) 
+    # Option 2: Target pods of specific nodes (testing mixed node types)
+    # node_names: 
+    #   - ip-10-0-31-8.us-east-2.compute.internal      # Worker node 1
+    #   - ip-10-0-48-188.us-east-2.compute.internal    # Worker node 2  
+    #   - ip-10-0-14-59.us-east-2.compute.internal     # Master node 1