Add exclude_label functionality to pod disruption scenarios (#910)

* kill pod exclude label

Signed-off-by: Tullio Sebastiani <tsebasti@redhat.com>

* config alignment

Signed-off-by: Tullio Sebastiani <tsebasti@redhat.com>

---------

Signed-off-by: Tullio Sebastiani <tsebasti@redhat.com>

krkn/scenario_plugins/pod_disruption/models/models.py

@@ -13,6 +13,7 @@ class InputParams:
             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 []
+            self.exclude_label = config["exclude_label"] if "exclude_label" in config else ""
 
     namespace_pattern: str
     krkn_pod_recovery_time: int
@@ -22,4 +23,5 @@ class InputParams:
     label_selector: str
     name_pattern: str
     node_label_selector: str
-    node_names: list
+    node_names: list
+    exclude_label: str

krkn/scenario_plugins/pod_disruption/pod_disruption_scenario_plugin.py

@@ -191,6 +191,13 @@ class PodDisruptionScenarioPlugin(AbstractScenarioPlugin):
             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", node_label_selector=config.node_label_selector, node_names=config.node_names)
+        exclude_pods = set()
+        if config.exclude_label:
+            _exclude_pods = self.get_pods("",config.exclude_label,config.namespace_pattern, kubecli, field_selector="status.phase=Running", node_label_selector=config.node_label_selector, node_names=config.node_names)
+            for pod in _exclude_pods:
+                exclude_pods.add(pod[0])
+
+
         pods_count = len(pods)
         if len(pods) < config.kill:
             logging.error("Not enough pods match the criteria, expected {} but found only {} pods".format(
@@ -201,8 +208,11 @@ class PodDisruptionScenarioPlugin(AbstractScenarioPlugin):
         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])
+            if pod[0] in exclude_pods:
+                logging.info(f"Excluding {pod[0]} from chaos")
+            else:
+                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, config.node_label_selector, config.node_names)
         return 0

scenarios/openshift/etcd.yml

@@ -4,3 +4,4 @@
     namespace_pattern: ^openshift-etcd$
     label_selector: k8s-app=etcd
     krkn_pod_recovery_time: 120
+    exclude_label: "" # excludes pods marked with this label from chaos

scenarios/openshift/openshift-apiserver.yml

@@ -4,4 +4,5 @@
     namespace_pattern: ^openshift-apiserver$
     label_selector: app=openshift-apiserver-a
     krkn_pod_recovery_time: 120
+    exclude_label: "" # excludes pods marked with this label from chaos
 

scenarios/openshift/openshift-kube-apiserver.yml

@@ -4,4 +4,5 @@
     namespace_pattern: ^openshift-kube-apiserver$
     label_selector: app=openshift-kube-apiserver
     krkn_pod_recovery_time: 120
+    exclude_label: "" # excludes pods marked with this label from chaos
 

scenarios/openshift/prom_kill.yml

@@ -2,4 +2,5 @@
   config:
     namespace_pattern: ^openshift-monitoring$
     label_selector: statefulset.kubernetes.io/pod-name=prometheus-k8s-0
-    krkn_pod_recovery_time: 120
+    krkn_pod_recovery_time: 120
+    exclude_label: "" # excludes pods marked with this label from chaos

scenarios/openshift/regex_openshift_pod_kill.yml

@@ -5,3 +5,4 @@
     name_pattern: .*
     kill: 3
     krkn_pod_recovery_time: 120
+    exclude_label: "" # excludes pods marked with this label from chaos