fix(query-performance): Make clickhouse_lag celery metric cheap (#13397)

* Limit what we measure lag on * fix(query-performance): make clickhouse_lag faster for other tables
PostHog · Dec 19, 2022 · a86c122 · a86c122
1 parent 26c5246
commit a86c122
Show file tree

Hide file tree

Showing 7 changed files with 69 additions and 40 deletions.
diff --git a/posthog/async_migrations/migrations/0008_speed_up_kafka_timestamp_filters.py b/posthog/async_migrations/migrations/0008_speed_up_kafka_timestamp_filters.py
@@ -1,3 +1,5 @@
+from functools import cached_property
+
 import structlog
 from django.conf import settings
 
@@ -31,35 +33,40 @@ def get_table_definition(self) -> str:
         )
         return result[0][0] if len(result) > 0 else ""
 
-    operations = [
-        AsyncMigrationOperationSQL(
-            database=AnalyticsDBMS.CLICKHOUSE,
-            sql=f"ALTER TABLE sharded_events ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' ADD PROJECTION fast_max_kafka_timestamp (SELECT max(_timestamp))",
-            rollback=f"ALTER TABLE sharded_events ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' DROP PROJECTION fast_max_kafka_timestamp",
-        ),
-        AsyncMigrationOperationSQL(
-            database=AnalyticsDBMS.CLICKHOUSE,
-            sql=f"ALTER TABLE sharded_events ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' MATERIALIZE PROJECTION fast_max_kafka_timestamp",
-            rollback=None,
-        ),
-        AsyncMigrationOperationSQL(
-            database=AnalyticsDBMS.CLICKHOUSE,
-            sql=f"ALTER TABLE sharded_events ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' ADD INDEX kafka_timestamp_minmax _timestamp TYPE minmax GRANULARITY 3",
-            rollback=f"ALTER TABLE sharded_events ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' DROP INDEX kafka_timestamp_minmax",
-        ),
-        AsyncMigrationOperationSQL(
-            database=AnalyticsDBMS.CLICKHOUSE,
-            sql=f"ALTER TABLE sharded_events ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' MATERIALIZE INDEX kafka_timestamp_minmax",
-            rollback=None,
-        ),
-        AsyncMigrationOperationSQL(
-            database=AnalyticsDBMS.CLICKHOUSE,
-            sql=f"ALTER TABLE events_dead_letter_queue ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' ADD INDEX kafka_timestamp_minmax_dlq _timestamp TYPE minmax GRANULARITY 3",
-            rollback=f"ALTER TABLE events_dead_letter_queue ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' DROP INDEX kafka_timestamp_minmax_dlq",
-        ),
-        AsyncMigrationOperationSQL(
-            database=AnalyticsDBMS.CLICKHOUSE,
-            sql=f"ALTER TABLE events_dead_letter_queue ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' MATERIALIZE INDEX kafka_timestamp_minmax_dlq",
-            rollback=None,
-        ),
-    ]
+    @cached_property
+    def operations(self):
+        operations = []
+
+        for table in ["sharded_events", "person", "person_distinct_id2", "sharded_session_recording_events"]:
+            operations.extend(
+                [
+                    AsyncMigrationOperationSQL(
+                        database=AnalyticsDBMS.CLICKHOUSE,
+                        sql=f"ALTER TABLE {table} ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' ADD PROJECTION fast_max_kafka_timestamp_{table} (SELECT max(_timestamp))",
+                        rollback=f"ALTER TABLE {table} ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' DROP PROJECTION fast_max_kafka_timestamp_{table}",
+                    ),
+                    AsyncMigrationOperationSQL(
+                        database=AnalyticsDBMS.CLICKHOUSE,
+                        sql=f"ALTER TABLE {table} ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' MATERIALIZE PROJECTION fast_max_kafka_timestamp_{table}",
+                        rollback=None,
+                    ),
+                ]
+            )
+
+        for table in ["sharded_events", "events_dead_letter_queue"]:
+            operations.extend(
+                [
+                    AsyncMigrationOperationSQL(
+                        database=AnalyticsDBMS.CLICKHOUSE,
+                        sql=f"ALTER TABLE {table} ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' ADD INDEX kafka_timestamp_minmax_{table} _timestamp TYPE minmax GRANULARITY 3",
+                        rollback=f"ALTER TABLE {table} ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' DROP INDEX kafka_timestamp_minmax_{table}",
+                    ),
+                    AsyncMigrationOperationSQL(
+                        database=AnalyticsDBMS.CLICKHOUSE,
+                        sql=f"ALTER TABLE {table} ON CLUSTER '{settings.CLICKHOUSE_CLUSTER}' MATERIALIZE INDEX kafka_timestamp_minmax_{table}",
+                        rollback=None,
+                    ),
+                ]
+            )
+
+        return operations
diff --git a/posthog/celery.py b/posthog/celery.py
@@ -293,7 +293,7 @@ def pg_plugin_server_query_timing():
             pass
 
 
-CLICKHOUSE_TABLES = ["events", "person", "person_distinct_id", "person_distinct_id2", "session_recording_events"]
+CLICKHOUSE_TABLES = ["events", "person", "person_distinct_id2", "session_recording_events"]
 
 
 @app.task(ignore_result=True)

diff --git a/posthog/clickhouse/dead_letter_queue.py b/posthog/clickhouse/dead_letter_queue.py
@@ -1,3 +1,4 @@
+from posthog.clickhouse.indexes import index_by_kafka_timestamp
 from posthog.clickhouse.kafka_engine import KAFKA_COLUMNS, kafka_engine, ttl_period
 from posthog.clickhouse.table_engines import ReplacingMergeTree
 from posthog.kafka_client.topics import KAFKA_DEAD_LETTER_QUEUE
@@ -43,7 +44,7 @@
     cluster=CLICKHOUSE_CLUSTER,
     extra_fields=f"""
     {KAFKA_COLUMNS}
-    , INDEX kafka_timestamp_minmax_dlq _timestamp TYPE minmax GRANULARITY 3
+    , {index_by_kafka_timestamp(DEAD_LETTER_QUEUE_TABLE)}
     """,
     engine=DEAD_LETTER_QUEUE_TABLE_ENGINE(),
     ttl_period=ttl_period("_timestamp", 4),  # 4 weeks

diff --git a/posthog/clickhouse/indexes.py b/posthog/clickhouse/indexes.py
@@ -0,0 +1,8 @@
+# Speeds up selecting max(_timestamp)
+def projection_for_max_kafka_timestamp(table: str):
+    return f"PROJECTION fast_max_kafka_timestamp_{table} (SELECT max(_timestamp))"
+
+
+# Speeds up filtering by _timestamp columns
+def index_by_kafka_timestamp(table: str):
+    return f"INDEX kafka_timestamp_minmax_{table} _timestamp TYPE minmax GRANULARITY 3"
diff --git a/posthog/models/event/sql.py b/posthog/models/event/sql.py
@@ -1,6 +1,7 @@
 from django.conf import settings
 
 from posthog.clickhouse.base_sql import COPY_ROWS_BETWEEN_TEAMS_BASE_SQL
+from posthog.clickhouse.indexes import index_by_kafka_timestamp, projection_for_max_kafka_timestamp
 from posthog.clickhouse.kafka_engine import KAFKA_COLUMNS, STORAGE_POLICY, kafka_engine, trim_quotes_expr
 from posthog.clickhouse.table_engines import Distributed, ReplacingMergeTree, ReplicationScheme
 from posthog.kafka_client.topics import KAFKA_EVENTS_JSON
@@ -79,9 +80,9 @@
     engine=EVENTS_DATA_TABLE_ENGINE(),
     extra_fields=KAFKA_COLUMNS,
     materialized_columns=EVENTS_TABLE_MATERIALIZED_COLUMNS,
-    indexes="""
-    , PROJECTION fast_max_kafka_timestamp (SELECT max(_timestamp))
-    , INDEX kafka_timestamp_minmax _timestamp TYPE minmax GRANULARITY 3
+    indexes=f"""
+    , {projection_for_max_kafka_timestamp(EVENTS_DATA_TABLE())}
+    , {index_by_kafka_timestamp(EVENTS_DATA_TABLE())}
     """,
     sample_by="SAMPLE BY cityHash64(distinct_id)",
     storage_policy=STORAGE_POLICY(),

diff --git a/posthog/models/person/sql.py b/posthog/models/person/sql.py
@@ -1,4 +1,5 @@
 from posthog.clickhouse.base_sql import COPY_ROWS_BETWEEN_TEAMS_BASE_SQL
+from posthog.clickhouse.indexes import index_by_kafka_timestamp
 from posthog.clickhouse.kafka_engine import KAFKA_COLUMNS, STORAGE_POLICY, kafka_engine
 from posthog.clickhouse.table_engines import CollapsingMergeTree, ReplacingMergeTree
 from posthog.kafka_client.topics import KAFKA_PERSON, KAFKA_PERSON_DISTINCT_ID, KAFKA_PERSON_UNIQUE_ID
@@ -39,7 +40,10 @@
     table_name=PERSONS_TABLE,
     cluster=CLICKHOUSE_CLUSTER,
     engine=PERSONS_TABLE_ENGINE(),
-    extra_fields=KAFKA_COLUMNS,
+    extra_fields=f"""
+    {KAFKA_COLUMNS}
+    , {index_by_kafka_timestamp(PERSONS_TABLE)}
+    """,
     storage_policy=STORAGE_POLICY(),
 )
 
@@ -182,7 +186,11 @@
     table_name=PERSON_DISTINCT_ID2_TABLE,
     cluster=CLICKHOUSE_CLUSTER,
     engine=PERSON_DISTINCT_ID2_TABLE_ENGINE(),
-    extra_fields=KAFKA_COLUMNS + "\n, _partition UInt64",
+    extra_fields=f"""
+    {KAFKA_COLUMNS}
+    , _partition UInt64
+    , {index_by_kafka_timestamp(PERSON_DISTINCT_ID2_TABLE)}
+    """,
 )
 
 KAFKA_PERSON_DISTINCT_ID2_TABLE_SQL = lambda: PERSON_DISTINCT_ID2_TABLE_BASE_SQL.format(

diff --git a/posthog/models/session_recording_event/sql.py b/posthog/models/session_recording_event/sql.py
@@ -1,5 +1,6 @@
 from django.conf import settings
 
+from posthog.clickhouse.indexes import index_by_kafka_timestamp
 from posthog.clickhouse.kafka_engine import KAFKA_COLUMNS, kafka_engine, ttl_period
 from posthog.clickhouse.table_engines import Distributed, ReplacingMergeTree, ReplicationScheme
 from posthog.kafka_client.topics import KAFKA_SESSION_RECORDING_EVENTS
@@ -86,7 +87,10 @@
     table_name=SESSION_RECORDING_EVENTS_DATA_TABLE(),
     cluster=settings.CLICKHOUSE_CLUSTER,
     materialized_columns=SESSION_RECORDING_EVENTS_MATERIALIZED_COLUMNS,
-    extra_fields=KAFKA_COLUMNS,
+    extra_fields=f"""
+    {KAFKA_COLUMNS}
+    , {index_by_kafka_timestamp(SESSION_RECORDING_EVENTS_DATA_TABLE())}
+    """,
     engine=SESSION_RECORDING_EVENTS_DATA_TABLE_ENGINE(),
     ttl_period=ttl_period(),
 )