Code import - branch release/SIENTIAPDE-1645

model_manager/metrics.py

@@ -0,0 +1,110 @@
+"""
+Model Manager Metrics Module
+
+This module defines all Prometheus metrics used by the Sientia DataOps Model Manager system
+for monitoring and observability. The metrics provide insights into system performance,
+training operations, and operational health.
+
+The metrics are designed to be scraped by Prometheus and can be visualized in
+Grafana or other monitoring dashboards to provide real-time visibility into
+the system's operation.
+
+Key Metric Categories:
+- Application Health: Overall system status and availability
+
+Metric Labels:
+- pod_id: Kubernetes pod identifier for multi-instance deployments
+"""
+
+from prometheus_client import Counter, Gauge, Histogram
+
+# Application health metric
+APP_UP = Gauge(
+    'app_up',
+    'Indicates if the application is running (1) or shutting down (0)',
+    ['pod_id'],
+)
+
+_TRAINING_LABELS = ['pod_id', 'model_name', 'model_type']
+
+SIENTIA_TRAINING_INFO = Gauge(
+    'sientia_training_info',
+    'Metadata and execution timestamp (ms) of the last successful model training run',
+    [
+        'pod_id',
+        'model_name',
+        'model_type',
+        'dataset_train_rows',
+        'dataset_val_rows',
+        'feature_count',
+        'mse',
+        'mae',
+        'r2',
+    ],
+)
+
+SIENTIA_TRAINING_MODEL_TRAINED_TOTAL = Counter(
+    'sientia_training_model_trained_total',
+    'Number of successfully completed model training runs',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_DATA_PREPARATION_LAG = Histogram(
+    'sientia_training_data_preparation_lag',
+    'Latency of prepare_training_data() in seconds',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_DATA_PREPARATION_ERROR_COUNT_TOTAL = Counter(
+    'sientia_training_data_preparation_error_count_total',
+    'Number of failures in prepare_training_data()',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_MODEL_FIT_LAG = Histogram(
+    'sientia_training_model_fit_lag',
+    'Latency of wrapper.train() (model fitting) in seconds',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_MODEL_FIT_ERROR_COUNT_TOTAL = Counter(
+    'sientia_training_model_fit_error_count_total',
+    'Number of failures in wrapper.train() (model fitting)',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_MODEL_QUALITY_MSE = Gauge(
+    'sientia_training_model_quality_mse',
+    'Mean Squared Error of the last successful model training',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_MODEL_QUALITY_MAE = Gauge(
+    'sientia_training_model_quality_mae',
+    'Mean Absolute Error of the last successful model training',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_MODEL_QUALITY_R2 = Gauge(
+    'sientia_training_model_quality_r2',
+    'R-squared of the last successful model training',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_DATASET_TRAIN_ROWS = Gauge(
+    'sientia_training_dataset_train_rows',
+    'Number of rows in the training dataset after preparation',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_DATASET_VAL_ROWS = Gauge(
+    'sientia_training_dataset_val_rows',
+    'Number of rows in the validation dataset after preparation',
+    _TRAINING_LABELS,
+)
+
+SIENTIA_TRAINING_FEATURE_COUNT = Gauge(
+    'sientia_training_feature_count',
+    'Number of input feature columns used for training',
+    _TRAINING_LABELS,
+)