test/persist: add test that compaction is happening as expected

test/persistence/compaction/compaction.td

@@ -0,0 +1,164 @@
+# Copyright Materialize, Inc. and contributors. All rights reserved.
+#
+# Use of this software is governed by the Business Source License
+# included in the LICENSE file at the root of this repository.
+#
+# As of the Change Date specified in that file, in accordance with
+# the Business Source License, use of this software will be governed
+# by the Apache License, Version 2.0.
+
+#
+# Smoke tests that verify compaction occurs as expected.
+#
+
+$ set-sql-timeout duration=30s
+
+> CREATE MATERIALIZED VIEW most_recent_mz_metrics AS
+  SELECT * FROM
+  (SELECT DISTINCT metric FROM mz_metrics) grp,
+  LATERAL (
+    SELECT value FROM mz_metrics
+    WHERE metric = grp.metric
+    ORDER BY time DESC LIMIT 1)
+
+# At the beginning there are no trace batches.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+0
+
+> CREATE TABLE compaction (f1 INTEGER);
+
+> INSERT INTO compaction VALUES (1);
+
+# The insert should make it into trace, and form 1 trace batch.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+1
+
+# Read after write to coerce different writes to happen at different timestamps.
+> SELECT * FROM compaction
+f1
+----
+1
+
+> INSERT INTO compaction VALUES (2);
+
+> SELECT * FROM compaction
+f1
+----
+1
+2
+
+# The previous insert should get compacted into one batch with the one before it.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+1
+
+> INSERT INTO compaction VALUES (3);
+
+> SELECT * FROM compaction
+f1
+----
+1
+2
+3
+
+# The most recent insert forms a new batch because it is at a lower compaction level.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+2
+
+> INSERT INTO compaction VALUES (4);
+
+> SELECT * FROM compaction
+f1
+----
+1
+2
+3
+4
+
+# The previous insert should compact together with the previous batch, which then compacts recursively
+# up, leaving just one batch.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+1
+
+> INSERT INTO compaction VALUES (5);
+
+> SELECT * FROM compaction
+f1
+----
+1
+2
+3
+4
+5
+
+# The previous insert forms a new batch.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+2
+
+> INSERT INTO compaction VALUES (6);
+
+> SELECT * FROM compaction
+f1
+----
+1
+2
+3
+4
+5
+6
+
+# The previous insert gets merged together with the one just before it.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+2
+
+> INSERT INTO compaction VALUES (7);
+
+> SELECT * FROM compaction
+f1
+----
+1
+2
+3
+4
+5
+6
+7
+
+# The previous insert forms a new batch.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+3
+
+> INSERT INTO compaction VALUES (8);
+
+> SELECT * FROM compaction
+f1
+----
+1
+2
+3
+4
+5
+6
+7
+8
+
+# All of the previously inserted batches get compacted together into one batch.
+> SELECT value FROM most_recent_mz_metrics where metric = 'mz_persist_trace_blob_count'
+value
+-----
+1

test/persistence/mzcompose.py

@@ -31,6 +31,10 @@
     options=f"{mz_options} --disable-user-indexes",
 )
 
+mz_fast_metrics = Materialized(
+    options=f"{mz_options} --metrics-scraping-interval=1s",
+)
+
 prerequisites = ["zookeeper", "kafka", "schema-registry"]
 
 SERVICES = [
@@ -54,6 +58,7 @@ def workflow_default(c: Composition) -> None:
             workflow_failpoints(c)
 
     workflow_disable_user_indexes(c)
+    workflow_compaction(c)
 
 
 def workflow_kafka_sources(c: Composition) -> None:
@@ -149,3 +154,17 @@ def workflow_disable_user_indexes(c: Composition) -> None:
     c.rm("materialized", "testdrive-svc", destroy_volumes=True)
 
     c.rm_volumes("mzdata")
+
+
+def workflow_compaction(c: Composition) -> None:
+    with c.override(mz_fast_metrics):
+        c.up("materialized")
+        c.wait_for_materialized()
+
+        c.run("testdrive-svc", "compaction/compaction.td")
+
+        c.kill("materialized")
+
+    c.rm("materialized", "testdrive-svc", destroy_volumes=True)
+
+    c.rm_volumes("mzdata")