From af8f0df2b0ad66dad1983c2d5a41b5e49fd7ebfb Mon Sep 17 00:00:00 2001
From: Tullio Sebastiani <tsebasti@redhat.com>
Date: Wed, 2 Aug 2023 12:22:56 +0200
Subject: [PATCH] prometheus telemetry upload + config.yaml

some fixes


typos and logs


max retries in config


telemetry id with run_uuid


safe_logger
---
 config/config.yaml |  7 ++++++-
 run_kraken.py      | 45 ++++++++++++++++++++++++++++++++-------------
 2 files changed, 38 insertions(+), 14 deletions(-)

diff --git a/config/config.yaml b/config/config.yaml
index 088946bc8..d2dfb0141 100644
--- a/config/config.yaml
+++ b/config/config.yaml
@@ -68,4 +68,9 @@ telemetry:
     enabled: True
     api_url: https://ulnmf9xv7j.execute-api.us-west-2.amazonaws.com/production
     username: username
-    password: password
\ No newline at end of file
+    password: password
+    prometheus_backup: True
+    full_prometheus_backup: False
+    backup_threads: 5
+    archive_path: /tmp
+    max_retries: 0
\ No newline at end of file
diff --git a/run_kraken.py b/run_kraken.py
index 76b7c81bd..11538e1b5 100644
--- a/run_kraken.py
+++ b/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, KrknTelemetry, ChaosRunTelemetry
+from krkn_lib_kubernetes import KrknLibKubernetes, KrknTelemetry, ChaosRunTelemetry, SafeLogger
 
 KUBE_BURNER_URL = (
     "https://github.com/cloud-bulldozer/kube-burner/"
@@ -98,15 +98,32 @@ def main(cfg):
             )
             sys.exit(1)
         logging.info("Initializing client to talk to the Kubernetes cluster")
+
+        # Generate uuid for the run
+        if run_uuid:
+            logging.info(
+                "Using the uuid defined by the user for the run: %s" % run_uuid
+            )
+        else:
+            run_uuid = str(uuid.uuid4())
+            logging.info("Generated a uuid for the run: %s" % run_uuid)
+
+        # request_id for telemetry is generated once here and used everywhere
+        telemetry_request_id = f"{int(time.time())}-{run_uuid}"
+        telemetry_log_file = f'{config["telemetry"]["archive_path"]}/{telemetry_request_id}.log'
+        safe_logger = SafeLogger(filename=telemetry_log_file)
+
         try:
             kubeconfig_path
             os.environ["KUBECONFIG"] = str(kubeconfig_path)
+            # krkn-lib-kubernetes init
             kubecli = KrknLibKubernetes(kubeconfig_path=kubeconfig_path)
-            telemetry = KrknTelemetry()
-
         except NameError:
             kubecli.initialize_clients(None)
 
+        # KrknTelemetry init
+        telemetry = KrknTelemetry(safe_logger, kubecli)
+
         # find node kraken might be running on
         kubecli.find_kraken_node()
 
@@ -143,14 +160,7 @@ def main(cfg):
         if deploy_performance_dashboards:
             performance_dashboards.setup(dashboard_repo, distribution)
 
-        # Generate uuid for the run
-        if run_uuid:
-            logging.info(
-                "Using the uuid defined by the user for the run: %s" % run_uuid
-            )
-        else:
-            run_uuid = str(uuid.uuid4())
-            logging.info("Generated a uuid for the run: %s" % run_uuid)
+
 
         # Initialize the start iteration to 0
         iteration = 0
@@ -174,6 +184,7 @@ def main(cfg):
         start_time = int(time.time())
         litmus_installed = False
         chaos_telemetry = ChaosRunTelemetry()
+        chaos_telemetry.run_uuid = run_uuid
         # Loop to run the chaos starts here
         while int(iteration) < iterations and run_signal != "STOP":
             # Inject chaos scenarios specified in the config
@@ -364,9 +375,17 @@ def main(cfg):
 
             iteration += 1
             logging.info("")
-        # send telemetry
 
-        telemetry.send_telemetry(config["telemetry"],str(uuid.uuid1()), chaos_telemetry, kubecli)
+        # telemetry
+        logging.info(f"telemetry data will be stored on s3 bucket folder: {telemetry_request_id}")
+        logging.info(f"telemetry upload log: {safe_logger.log_file_name}")
+
+        telemetry.send_telemetry(config["telemetry"], telemetry_request_id, chaos_telemetry)
+        safe_logger.info("archives download started:")
+        prometheus_archive_files = telemetry.get_ocp_prometheus_data(config["telemetry"], telemetry_request_id)
+        safe_logger.info("archives upload started:")
+        telemetry.put_ocp_prometheus_data(config["telemetry"], prometheus_archive_files, telemetry_request_id)
+
         # Capture the end time
         end_time = int(time.time())