run_kraken.py + arcaflow scenarios refactoring

kraken/arcaflow_plugin/arcaflow_plugin.py

@@ -1,3 +1,5 @@
+import time
+
 import arcaflow
 import os
 import yaml
@@ -6,22 +8,31 @@
 from pathlib import Path
 from typing import List
 from .context_auth import ContextAuth
+from krkn_lib_kubernetes import ScenarioTelemetry, KrknTelemetry
 
 
-def run(scenarios_list: List[str], kubeconfig_path: str):
+def run(scenarios_list: List[str], kubeconfig_path: str, telemetry: KrknTelemetry) -> (list[str],list[ScenarioTelemetry]):
+    scenario_telemetries: list[ScenarioTelemetry] = []
+    failed_post_scenarios = []
     for scenario in scenarios_list:
+        scenario_telemetry = ScenarioTelemetry()
+        scenario_telemetry.scenario = scenario
+        scenario_telemetry.startTimeStamp = time.time()
+        telemetry.set_parameters_base64(scenario_telemetry,scenario)
         engine_args = build_args(scenario)
-        run_workflow(engine_args, kubeconfig_path)
+        status_code = run_workflow(engine_args, kubeconfig_path)
+        scenario_telemetry.endTimeStamp = time.time()
+        scenario_telemetry.exitStatus = status_code
+        scenario_telemetries.append(scenario_telemetry)
+        if status_code != 0:
+            failed_post_scenarios.append(scenario)
+    return failed_post_scenarios, scenario_telemetries
 
 
-def run_workflow(engine_args: arcaflow.EngineArgs, kubeconfig_path: str):
+def run_workflow(engine_args: arcaflow.EngineArgs, kubeconfig_path: str) -> int:
     set_arca_kubeconfig(engine_args, kubeconfig_path)
     exit_status = arcaflow.run(engine_args)
-    if exit_status != 0:
-        logging.error(
-            f"failed to run arcaflow scenario {engine_args.input}"
-        )
-        sys.exit(exit_status)
+    return exit_status
 
 
 def build_args(input_file: str) -> arcaflow.EngineArgs:

run_kraken.py

@@ -25,7 +25,7 @@
 import server as server
 import kraken.prometheus.client as promcli
 from kraken import plugins
-from krkn_lib_kubernetes import KrknLibKubernetes
+from krkn_lib_kubernetes import KrknLibKubernetes, KrknTelemetry, ChaosRunTelemetry
 
 KUBE_BURNER_URL = (
     "https://github.com/cloud-bulldozer/kube-burner/"
@@ -102,6 +102,8 @@ def main(cfg):
             kubeconfig_path
             os.environ["KUBECONFIG"] = str(kubeconfig_path)
             kubecli = KrknLibKubernetes(kubeconfig_path=kubeconfig_path)
+            telemetry = KrknTelemetry()
+
         except NameError:
             kubecli.initialize_clients(None)
 
@@ -174,7 +176,7 @@ def main(cfg):
         # Capture the start time
         start_time = int(time.time())
         litmus_installed = False
-
+        chaos_telemetry = ChaosRunTelemetry()
         # Loop to run the chaos starts here
         while int(iteration) < iterations and run_signal != "STOP":
             # Inject chaos scenarios specified in the config
@@ -206,9 +208,10 @@ def main(cfg):
                             )
                             sys.exit(1)
                         elif scenario_type == "arcaflow_scenarios":
-                            failed_post_scenarios = arcaflow_plugin.run(
-                                scenarios_list, kubeconfig_path
+                            failed_post_scenarios, scenario_telemetries = arcaflow_plugin.run(
+                                scenarios_list, kubeconfig_path, telemetry
                             )
+                            chaos_telemetry.scenarios.extend(scenario_telemetries)
 
                         elif scenario_type == "plugin_scenarios":
                             failed_post_scenarios = plugins.run(
@@ -355,7 +358,8 @@ def main(cfg):
 
             iteration += 1
             logging.info("")
-
+        # send telemetry
+        telemetry.send_telemetry(config["telemetry"], chaos_telemetry)
         # Capture the end time
         end_time = int(time.time())