Merge branch 'master' into alexy_ck_job_6

Author: ayefimov-1

.zuul.yaml

@@ -138,7 +138,6 @@
 
 - job:
     name: feature-verification-tests-noop
-    parent: noop
     description: |
       A job that always passes. Runs when there's a change to jobs that don't
       need full zuul to run but still need to report a pass.
@@ -149,6 +148,7 @@
     name: functional-periodic-telemetry-with-ceph
     parent: podified-multinode-hci-deployment-crc-1comp-backends
     dependencies: ["telemetry-openstack-meta-content-provider-master"]
+    nodeset: telemetry-operator-all-the-features
     description: |
       Deploy OpenStack with Telemetry and Ceph. Run metric-verification with volume pool metric tests
     extra-vars: *functional_autoscaling_extra_vars
@@ -241,4 +241,3 @@
               # case the job definition changes.
               - ^roles/telemetry_verify_metrics/tasks/verify_ceilometer_volume_pool_metrics.yml$
               - ^.zuul.yaml$
-

roles/telemetry_chargeback/defaults/main.yml

@@ -7,10 +7,10 @@ cloudkitty_debug: false
 cloudkitty_debug_dir: "{{ (cloudkitty_debug | bool) | ternary(artifacts_dir_zuul + '/debug_ck_db', '') }}"
 
 # Directory paths
-logs_dir_zuul: "{{ ansible_env.HOME }}/ci-framework-data/logs"
-artifacts_dir_zuul: "{{ ansible_env.HOME }}/ci-framework-data/artifacts"
-cert_dir: "{{ ansible_user_dir }}/ck-certs"
-local_cert_dir: "{{ ansible_env.HOME }}/ci-framework-data/flush_certs"
+logs_dir_zuul: "{{ cifmw_basedir }}/logs"
+artifacts_dir_zuul: "{{ cifmw_basedir }}/artifacts"
+cert_dir: "{{ cifmw_basedir }}/ck-certs"
+local_cert_dir: "{{ cifmw_basedir }}/flush_certs"
 remote_cert_dir: "osp-certs"
 
 # Cloudkitty certificates and secrets

roles/telemetry_chargeback/files/gen_db_summary.py

@@ -118,11 +118,11 @@ def _apply_mutate(qty: float, mutate: str) -> float:
     elif mutate_upper == "FLOOR":
         return math.floor(qty)
     elif mutate_upper == "NUMBOOL":
-        # If qty equals 0, leave it at 0. Else, set it to 1.
+        # If qty near 0, set it at 0. Else, set it to 1.
         return 0.0 if abs(qty) < 1e-9 else 1.0
     elif mutate_upper == "NOTNUMBOOL":
-        # If qty equals 0, set it to 1. Else, set it to 0.
-        return 1.0 if qty == 0 else 0.0
+        # If qty near 0, set it to 1. Else, set it to 0.
+        return 1.0 if abs(qty) < 1e-9 else 0.0
     else:  # NONE or any unrecognized value
         return qty
 
@@ -220,6 +220,11 @@ def aggregate_rates_by_type(
 
 
 def build_summary(pairs: list[tuple[str, str]]) -> dict[str, Any]:
+    # Early exit if no pairs
+    if not pairs:
+        print("Error: No log entries to summarize", file=sys.stderr)
+        sys.exit(1)
+
     log_count = len(pairs)
     per_ts = Counter(ts for ts, _ in pairs)
     n_ts = len(per_ts)
@@ -228,31 +233,29 @@ def build_summary(pairs: list[tuple[str, str]]) -> dict[str, Any]:
     if counts and len(set(counts)) > 1:
         mps = "ERROR"
 
-    if pairs:
-        first = json.loads(pairs[0][1])
-        last = json.loads(pairs[-1][1])
-        time_block = {
-            "begin_step": {
-                "nanosec": int(pairs[0][0]),
-                "begin": first.get("start"),
-                "end": first.get("end"),
-            },
-            "end_step": {
-                "nanosec": int(pairs[-1][0]),
-                "begin": last.get("start"),
-                "end": last.get("end"),
-            },
-        }
-    else:
-        empty = {"nanosec": None, "begin": None, "end": None}
-        time_block = {"begin_step": empty.copy(), "end_step": empty.copy()}
+    # Parse first and last entries (guaranteed to exist after early exit check)
+    first = json.loads(pairs[0][1])
+    last = json.loads(pairs[-1][1])
+
+    time_block = {
+        "begin_step": {
+            "nanosec": int(pairs[0][0]),
+            "begin": first.get("start"),
+            "end": first.get("end"),
+        },
+        "end_step": {
+            "nanosec": int(pairs[-1][0]),
+            "begin": last.get("start"),
+            "end": last.get("end"),
+        },
+    }
 
     # Get aggregated data by type
     by_types, total_r, qty_by_types = aggregate_rates_by_type(pairs)
 
     # Get overall time range for by_type entries
-    begin_time = first.get("start") if pairs else None
-    end_time = last.get("end") if pairs else None
+    begin_time = first.get("start")
+    end_time = last.get("end")
 
     # Build flat list of entries
     rate_list = []
@@ -292,6 +295,31 @@ def write_yaml(path: Path, doc: dict[str, Any]) -> None:
         )
 
 
+def _str_to_bool(value: str) -> bool:
+    """
+    Convert string to boolean.
+
+    Args:
+        value: String representation of boolean.
+
+    Returns:
+        Boolean value.
+
+    Raises:
+        argparse.ArgumentTypeError: If value cannot be converted.
+    """
+    if isinstance(value, bool):
+        return value
+    if value.lower() in ('yes', 'true', 't', 'y', '1'):
+        return True
+    elif value.lower() in ('no', 'false', 'f', 'n', '0'):
+        return False
+    else:
+        raise argparse.ArgumentTypeError(
+            f"Boolean value expected. Got: {value}"
+        )
+
+
 def main() -> None:
     parser = argparse.ArgumentParser(
         description=(
@@ -311,10 +339,12 @@ def main() -> None:
     )
     parser.add_argument(
         "--debug",
-        action="store_true",
+        type=_str_to_bool,
+        default=False,
+        metavar="BOOL",
         help=(
             "Enable debug mode: write <stem>_diff.txt with one "
-            "[ts,log] JSON per line."
+            "[ts,log] JSON per line (true/false)."
         ),
     )
     parser.add_argument(
@@ -323,8 +353,7 @@ def main() -> None:
         default=None,
         metavar="DIR",
         help=(
-            "Directory for debug output. If not specified, uses the "
-            "directory from -o output path."
+            "Directory for debug output. Required when --debug is enabled."
         ),
     )
     args = parser.parse_args()
@@ -338,9 +367,14 @@ def main() -> None:
     pairs = extract_and_sort(args.json)
 
     if args.debug:
-        # Determine debug directory: use --debug_dir if provided,
-        # otherwise use output directory
-        debug_dir = args.debug_dir if args.debug_dir else out_path.parent
+        # Require debug directory when debug mode is enabled
+        if not args.debug_dir:
+            print(
+                "Error: --debug_dir is required when --debug is enabled",
+                file=sys.stderr
+            )
+            sys.exit(1)
+        debug_dir = args.debug_dir
         debug_dir.mkdir(parents=True, exist_ok=True)
         dbg_file = debug_dir / f"{args.json.stem}_diff.txt"
         with dbg_file.open("w", encoding="utf-8") as f:
@@ -350,13 +384,6 @@ def main() -> None:
     doc = build_summary(pairs)
     write_yaml(out_path, doc)
 
-    if doc["data_summary"]["metrics_per_step"] == "ERROR":
-        per_ts = Counter(ts for ts, _ in pairs)
-        exp = next(iter(per_ts.values()), 0)
-        for ts in sorted(per_ts, key=int):
-            if per_ts[ts] != exp:
-                print(ts, per_ts[ts], file=sys.stdout)
-
 
 if __name__ == "__main__":
     main()

roles/telemetry_chargeback/files/gen_synth_loki_data.py

@@ -109,6 +109,7 @@ def generate_loki_data(
     end_time: datetime,
     time_step_seconds: int,
     config: Dict[str, Any],
+    scenario_name: str,
     reverse_timestamps: bool = True,
 ):
     """
@@ -121,6 +122,7 @@ def generate_loki_data(
         end_time (datetime): The end time for data generation.
         time_step_seconds (int): The duration of each log entry in seconds.
         config (Dict[str, Any]): Configuration dictionary loaded from file.
+        scenario_name (str): Name of the test scenario for Loki labeling.
         reverse_timestamps (bool): If True, sort timestamps in descending order
             (newest first, oldest last). If False, sort in ascending order
             (oldest first, newest last). Default is True (descending).
@@ -138,6 +140,7 @@ def generate_loki_data(
     logger.debug(f"Time range in epoch seconds: {start_epoch} to {end_epoch}")
 
     log_data_list = []  # This list will hold all our data points
+    last_end_of_step_epoch = None  # Track last entry's end epoch
 
     # Loop through the time range and generate data points
     for current_epoch in range(
@@ -164,13 +167,13 @@ def generate_loki_data(
             "end_time": end_str
         })
 
+        # Track the last end epoch
+        last_end_of_step_epoch = end_of_step_epoch
+
     # Add final entry that ends at end_epoch (current time)
-    if log_data_list and end_epoch > start_epoch:
+    if log_data_list and end_epoch > start_epoch and last_end_of_step_epoch:
         # Calculate start of final entry based on end of last generated entry
-        last_entry_end = log_data_list[-1]["end_time"]
-        # Parse the last entry's end time to get the epoch
-        last_end_dt = datetime.fromisoformat(last_entry_end)
-        final_start_epoch = int(last_end_dt.timestamp()) + 1
+        final_start_epoch = last_end_of_step_epoch + 1
         final_nanoseconds = int(final_start_epoch * 1_000_000_000)
 
         # Only add if the final entry would have a valid duration
@@ -339,12 +342,16 @@ def tojson_preserve_order(obj):
 
         log_types_list.append(log_types_with_dates)
 
-    # Get loki_stream configuration
+    # Get loki_stream configuration and add scenario
     loki_stream = config.get("loki_stream", {})
     if not loki_stream:
         logger.warning("No loki_stream configuration found, using defaults")
         loki_stream = {"service": "cloudkitty"}
 
+    # Add scenario name to loki_stream labels
+    loki_stream["scenario"] = scenario_name
+    logger.info(f"Adding scenario label: {scenario_name}")
+
     # Build template context with generic log type information
     template_context = {
         "log_data": log_data_list,
@@ -379,6 +386,31 @@ def tojson_preserve_order(obj):
         sys.exit(1)
 
 
+def _str_to_bool(value: str) -> bool:
+    """
+    Convert string to boolean.
+
+    Args:
+        value: String representation of boolean.
+
+    Returns:
+        Boolean value.
+
+    Raises:
+        argparse.ArgumentTypeError: If value cannot be converted.
+    """
+    if isinstance(value, bool):
+        return value
+    if value.lower() in ('yes', 'true', 't', 'y', '1'):
+        return True
+    elif value.lower() in ('no', 'false', 'f', 'n', '0'):
+        return False
+    else:
+        raise argparse.ArgumentTypeError(
+            f"Boolean value expected. Got: {value}"
+        )
+
+
 def main():
     """Main entry point for the script."""
     parser = argparse.ArgumentParser(
@@ -421,8 +453,10 @@ def main():
     )
     parser.add_argument(
         "--debug",
-        action="store_true",
-        help="Enable debug level logging for verbose output."
+        type=_str_to_bool,
+        default=False,
+        metavar="BOOL",
+        help="Enable debug level logging for verbose output (true/false)."
     )
     args = parser.parse_args()
 
@@ -435,7 +469,11 @@ def main():
         config = load_config(args.test)
     except (FileNotFoundError, ValueError) as e:
         logger.critical(f"Failed to load config: {e}")
-        return
+        sys.exit(1)
+
+    # Derive scenario name from test file path
+    scenario_name = args.test.stem
+    logger.info(f"Derived scenario name from test file: {scenario_name}")
 
     # Get generation parameters from config
     generation_config = config.get("generation", {})
@@ -456,14 +494,17 @@ def main():
             end_time=end_time_utc,
             time_step_seconds=step_seconds,
             config=config,
+            scenario_name=scenario_name,
             reverse_timestamps=args.reverse,
         )
     except FileNotFoundError:
         logger.error(
             "Process aborted because the template file was not found."
         )
+        sys.exit(1)
     except Exception as e:
         logger.critical(f"A critical, unhandled error stopped the script: {e}")
+        sys.exit(1)
 
 
 if __name__ == "__main__":

roles/telemetry_chargeback/tasks/cleanup_ck.yml

@@ -1,4 +1,5 @@
 ---
+### cleans up after each test scenario
 - name: "Cleanup local certificates"
   ansible.builtin.file:
     path: "{{ local_cert_dir }}"