Add extra_streaming_tables_have_latest.sql verification (#196)

airflow/dags/polygonetl_airflow/build_verify_streaming_dag.py

@@ -18,6 +18,7 @@ def build_verify_streaming_dag(
         destination_dataset_project_id,
         chain='polygon',
         notification_emails=None,
+        extra_streaming_tables=None,
         start_date=datetime(2018, 7, 1),
         schedule='*/10 * * * *',
         max_lag_in_minutes=15):
@@ -51,16 +52,24 @@ def build_verify_streaming_dag(
 
     dags_folder = os.environ.get('DAGS_FOLDER', '/home/airflow/gcs/dags')
 
-    def add_verify_tasks(task, dependencies=None):
+    def add_verify_tasks(task, dependencies=None, params=None):
         # The queries in verify/sqls will fail when the condition is not met
         # Have to use this trick since the Python 2 version of BigQueryCheckOperator doesn't support standard SQL
         # and legacy SQL can't be used to query partitioned tables.
         sql_path = os.path.join(dags_folder, 'resources/stages/verify_streaming/sqls/{task}.sql'.format(task=task))
         sql = read_file(sql_path)
+
+        combined_params = environment.copy()
+        task_id = 'verify_{task}'.format(task=task)
+        if params:
+            combined_params.update(params)
+            serialized_params = '_'.join(params.values()).replace('.', '_')
+            task_id = task_id + '_' + serialized_params
+
         verify_task = BigQueryInsertJobOperator(
-            task_id=f"verify_{task}",
+            task_id=task_id,
             configuration={"query": {"query": sql, "useLegacySql": False}},
-            params=environment,
+            params=combined_params,
             dag=dag,
         )
         if dependencies is not None and len(dependencies) > 0:
@@ -78,6 +87,13 @@ def add_verify_tasks(task, dependencies=None):
 
     add_verify_tasks('transactions_count')
 
+    # Use this to verify the lag of a streaming job https://github.com/blockchain-etl/blockchain-etl-streaming by piping a Pub/Sub topic to a BigQuery Table
+    # https://cloud.google.com/blog/products/data-analytics/pub-sub-launches-direct-path-to-bigquery-for-streaming-analytics
+    if extra_streaming_tables is not None and len(extra_streaming_tables) > 0:
+        streaming_table_list = [table.strip() for table in extra_streaming_tables.split(',')]
+        for streaming_table in streaming_table_list:
+            add_verify_tasks('extra_streaming_tables_have_latest', params={'streaming_table': streaming_table})
+
     return dag
 
 

airflow/dags/polygonetl_airflow/variables.py

@@ -112,6 +112,11 @@ def read_parse_dag_vars(var_prefix, **kwargs):
 def read_verify_streaming_dag_vars(var_prefix, **kwargs):
     vars = {
         'destination_dataset_project_id': read_var('destination_dataset_project_id', var_prefix, True, **kwargs),
+        # Comma separated list of fully qualified BigQuery table names e.g. myproject.streamers.blocks1,myproject.streamers.blocks2
+        # Each table must have a column data containing a JSON object with field timestamp containing unix seconds
+        # Used in combination with https://cloud.google.com/blog/products/data-analytics/pub-sub-launches-direct-path-to-bigquery-for-streaming-analytics
+        # to verify lag of streaming jobs that output to Pub/Sub
+        'extra_streaming_tables': read_var('extra_streaming_tables', var_prefix, False, **kwargs),
         'notification_emails': read_var('notification_emails', None, False, **kwargs),
     }
 

airflow/dags/resources/stages/verify_streaming/sqls/extra_streaming_tables_have_latest.sql

@@ -0,0 +1,9 @@
+select if(
+(
+select timestamp_diff(
+  current_timestamp(),
+  (select max(timestamp_seconds(cast(json_extract(data, '$.timestamp') AS INTEGER)))
+  from `{{params.streaming_table}}`),
+  MINUTE)
+) < {{params.max_lag_in_minutes}}, 1,
+cast((select 'Streaming table {{params.streaming_table}} is lagging by more than {{params.max_lag_in_minutes}} minutes') as INT64))