rust/neon
1
0
Fork 0
mirror of https://github.com/neondatabase/neon.git synced 2026-10-28 06:26:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

Analyze clickbench performance with explain plans and pg_stat_statements (#6161)

Browse Source 
## Problem

To understand differences in performance between neon, aurora and rds we
want to collect explain analyze plans and pg_stat_statements for
selected benchmarking runs

## Summary of changes

Add workflow input options to collect explain and pg_stat_statements for
benchmarking workflow

Co-authored-by: BodoBolero <bodobolero@gmail.com>
This commit is contained in:
Bodobolero
2023-12-19 11:44:25 +00:00
committed by GitHub
co-authored by BodoBolero
parent b701394d7a
commit 73d247c464
2 changed files with 74 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test_runner/performance/test_perf_olap.py
+49 -2
View File
@@ -17,6 +17,27 @@ class LabelledQuery:
query: str
# This must run before all tests in this module
# create extension pg_stat_statements if it does not exist
# and TEST_OLAP_COLLECT_PG_STAT_STATEMENTS is set to true (default false)
# Theoretically this could be in a module or session scope fixture,
# however the code depends on other fixtures that have function scope
@pytest.mark.skipif(
os.getenv("TEST_OLAP_COLLECT_PG_STAT_STATEMENTS", "false").lower() == "false",
reason="Skipping - Creating extension pg_stat_statements",
)
@pytest.mark.remote_cluster
def test_clickbench_create_pg_stat_statements(remote_compare: RemoteCompare):
log.info("Creating extension pg_stat_statements")
query = LabelledQuery(
"Q_CREATE_EXTENSION", r"CREATE EXTENSION IF NOT EXISTS pg_stat_statements;"
)
run_psql(remote_compare, query, times=1, explain=False)
log.info("Reset pg_stat_statements")
query = LabelledQuery("Q_RESET", r"SELECT pg_stat_statements_reset();")
run_psql(remote_compare, query, times=1, explain=False)
# A list of queries to run.
# Please do not alter the label for the query, as it is used to identify it.
# Labels for ClickBench queries match the labels in ClickBench reports
@@ -78,6 +99,8 @@ QUERIES: Tuple[LabelledQuery, ...] = (
# fmt: on
)
EXPLAIN_STRING: str = "EXPLAIN (ANALYZE, VERBOSE, BUFFERS, COSTS, SETTINGS, FORMAT JSON)"
def get_scale() -> List[str]:
# We parametrize each tpc-h and clickbench test with scale
@@ -88,7 +111,10 @@ def get_scale() -> List[str]:
return [scale]
def run_psql(env: RemoteCompare, labelled_query: LabelledQuery, times: int) -> None:
# run the query times times plus once with EXPLAIN VERBOSE if explain is requestd
def run_psql(
env: RemoteCompare, labelled_query: LabelledQuery, times: int, explain: bool = False
) -> None:
# prepare connstr:
# - cut out password from connstr to pass it via env
# - add options to connstr
@@ -108,6 +134,13 @@ def run_psql(env: RemoteCompare, labelled_query: LabelledQuery, times: int) -> N
log.info(f"Run {run}/{times}")
with env.zenbenchmark.record_duration(f"{label}/{run}"):
env.pg_bin.run_capture(["psql", connstr, "-c", query], env=environ)
if explain:
log.info(f"Explaining query {label}")
run += 1
with env.zenbenchmark.record_duration(f"{label}/EXPLAIN"):
env.pg_bin.run_capture(
["psql", connstr, "-c", f"{EXPLAIN_STRING} {query}"], env=environ
)
@pytest.mark.parametrize("scale", get_scale())
@@ -120,8 +153,9 @@ def test_clickbench(query: LabelledQuery, remote_compare: RemoteCompare, scale:
Based on https://github.com/ClickHouse/ClickBench/tree/c00135ca5b6a0d86fedcdbf998fdaa8ed85c1c3b/aurora-postgresql
The DB prepared manually in advance
"""
explain: bool = os.getenv("TEST_OLAP_COLLECT_EXPLAIN", "false").lower() == "true"
run_psql(remote_compare, query, times=3)
run_psql(remote_compare, query, times=3, explain=explain)
def tpch_queuies() -> Tuple[ParameterSet, ...]:
@@ -195,3 +229,16 @@ def test_user_examples(remote_compare: RemoteCompare):
""",
)
run_psql(remote_compare, query, times=3)
# This must run after all tests in this module
# Collect pg_stat_statements after running the tests if TEST_OLAP_COLLECT_PG_STAT_STATEMENTS is set to true (default false)
@pytest.mark.skipif(
os.getenv("TEST_OLAP_COLLECT_PG_STAT_STATEMENTS", "false").lower() == "false",
reason="Skipping - Collecting pg_stat_statements",
)
@pytest.mark.remote_cluster
def test_clickbench_collect_pg_stat_statements(remote_compare: RemoteCompare):
log.info("Collecting pg_stat_statements")
query = LabelledQuery("Q_COLLECT_PG_STAT_STATEMENTS", r"SELECT * from pg_stat_statements;")
run_psql(remote_compare, query, times=1, explain=False)