Skip to content

C
ClickHouse

2dot5 / ClickHouse

通知 3
Star 0
Fork 0
C
ClickHouse

体验新版 GitCode，发现更多精彩内容 >>

未验证 提交 d4445485 编写于 作者: A Alexander Kuzmenkov 提交者： GitHub
浏览文件
操作

Merge pull request #10028 from ClickHouse/aku/perftest-fixes 

Performance comparison fixes
上级 eea9845f 882e0d8b
展开全部 隐藏空白更改
内联 并排
Showing with 301 addition and 253 deletion
docker/test/performance-comparison/compare.sh
浏览文件 @ d4445485
......@@ -97,10 +97,6 @@ function run_tests
touch "$x"
done
# FIXME remove some broken long tests
rm "$test_prefix"/{IPv4,IPv6,modulo,parse_engine_file,number_formatting_formats,select_format}.xml ||:
test_files=$(ls "$test_prefix"/*.xml)
# FIXME a quick crutch to bring the run time down for the unstable tests --
# if some performance tests xmls were changed in a PR, run only these ones.
......@@ -126,6 +122,17 @@ function run_tests
test_files=$(ls "$test_prefix"/$CHPC_TEST_GLOB.xml)
fi
if [ "$test_files" == "" ]
then
# FIXME remove some broken long tests
for test_name in {IPv4,IPv6,modulo,parse_engine_file,number_formatting_formats,select_format,arithmetic,cryptographic_hashes,logical_functions_{medium,small}}
do
printf "$test_name\tMarked as broken (see compare.sh)\n" >> skipped-tests.tsv
rm "$test_prefix/$test_name.xml" ||:
done
test_files=$(ls "$test_prefix"/*.xml)
fi
# Run the tests.
test_name="<none>"
for test in $test_files
......@@ -275,9 +282,11 @@ create table test_times_tsv engine File(TSV, 'test-times.tsv') as
from test_time join wall_clock using test
order by avg_real_per_query desc;
create table all_queries_tsv engine File(TSV, 'all-queries.tsv') as
select left, right, diff, rd, test, query
from queries order by rd[3] desc;
create table all_tests_tsv engine File(TSV, 'all-queries.tsv') as
select left, right, diff,
floor(left > right ? left / right : right / left, 3),
rd, test, query
from queries order by test, query;
" 2> >(head -2 >> report-errors.rep) ||:
for version in {right,left}
......@@ -429,6 +438,7 @@ case "$stage" in
"report")
time report ||:
time "$script_dir/report.py" --report=all-queries > all-queries.html 2> >(head -2 >> report-errors.rep) ||:
time "$script_dir/report.py" > report.html
;&
esac
docker/test/performance-comparison/entrypoint.sh
浏览文件 @ d4445485
......@@ -90,17 +90,23 @@ export PYTHONIOENCODING=utf-8
# Use a default number of runs if not told otherwise
export CHPC_RUNS=${CHPC_RUNS:-7}
# By default, use the main comparison script from the tested package, so that we
# can change it in PRs.
script_path="right/scripts"
if [ -v CHPC_LOCAL_SCRIPT ]
then
script_path=".."
fi
# Even if we have some errors, try our best to save the logs.
set +e
# Use main comparison script from the tested package, so that we can change it
# in PRs.
# Older version use 'kill 0', so put the script into a separate process group
# FIXME remove set +m in April 2020
set +m
{ \
time ../download.sh "$REF_PR" "$REF_SHA" "$PR_TO_TEST" "$SHA_TO_TEST" && \
time stage=configure right/scripts/compare.sh ; \
time stage=configure "$script_path"/compare.sh ; \
} 2>&1 | ts "$(printf '%%Y-%%m-%%d %%H:%%M:%%S\t')" | tee compare.log
set -m
......
docker/test/performance-comparison/report.py
浏览文件 @ d4445485
#!/usr/bin/python3
import argparse
import ast
import collections
import csv
......@@ -8,6 +9,11 @@ import os
import sys
import traceback
parser = argparse.ArgumentParser(description='Create performance test report')
parser.add_argument('--report', default='main', choices=['main', 'all-queries'],
help='Which report to build')
args = parser.parse_args()
report_errors = []
error_tests = 0
slow_average_tests = 0
......@@ -16,7 +22,7 @@ slower_queries = 0
unstable_queries = 0
very_unstable_queries = 0
print("""
header_template = """
<!DOCTYPE html>
<html>
<style>
......@@ -56,7 +62,7 @@ tr:nth-child(odd) td {{filter: brightness(95%);}}
<div class="main">
<h1>ClickHouse performance comparison</h1>
""".format())
"""
table_anchor = 0
row_anchor = 0
......@@ -133,195 +139,252 @@ def printSimpleTable(caption, columns, rows):
print(tableRow(row))
print(tableEnd())
printSimpleTable('Tested commits', ['Old', 'New'],
[['<pre>{}</pre>'.format(x) for x in
[open('left-commit.txt').read(),
open('right-commit.txt').read()]]])
def print_changes():
rows = tsvRows('changed-perf.tsv')
if not rows:
return
global faster_queries, slower_queries
print(tableStart('Changes in performance'))
columns = [
'Old, s', # 0
'New, s', # 1
'Relative difference (new&nbsp;-&nbsp;old)/old', # 2
'Randomization distribution quantiles \
[5%,&nbsp;50%,&nbsp;95%,&nbsp;99%]', # 3
'Test', # 4
'Query', # 5
if args.report == 'main':
print(header_template.format())
printSimpleTable('Tested commits', ['Old', 'New'],
[['<pre>{}</pre>'.format(x) for x in
[open('left-commit.txt').read(),
open('right-commit.txt').read()]]])
def print_changes():
rows = tsvRows('changed-perf.tsv')
if not rows:
return
global faster_queries, slower_queries
print(tableStart('Changes in performance'))
columns = [
'Old, s', # 0
'New, s', # 1
'Relative difference (new&nbsp;-&nbsp;old)/old', # 2
'Randomization distribution quantiles \
[5%,&nbsp;50%,&nbsp;95%,&nbsp;99%]', # 3
'Test', # 4
'Query', # 5
]
print(tableHeader(columns))
attrs = ['' for c in columns]
for row in rows:
if float(row[2]) < 0.:
faster_queries += 1
attrs[2] = 'style="background: #adbdff"'
else:
slower_queries += 1
attrs[2] = 'style="background: #ffb0a0"'
print(tableRow(row, attrs))
print(tableEnd())
print_changes()
slow_on_client_rows = tsvRows('slow-on-client.tsv')
error_tests += len(slow_on_client_rows)
printSimpleTable('Slow on client',
['Client time, s', 'Server time, s', 'Ratio', 'Query'],
slow_on_client_rows)
def print_unstable_queries():
global unstable_queries
global very_unstable_queries
unstable_rows = tsvRows('unstable-queries.tsv')
if not unstable_rows:
return
unstable_queries += len(unstable_rows)
columns = [
'Old, s', #0
'New, s', #1
'Relative difference (new&nbsp;-&nbsp;old)/old', #2
'Randomization distribution quantiles [5%,&nbsp;50%,&nbsp;95%,&nbsp;99%]', #3
'Test', #4
'Query' #5
]
print(tableHeader(columns))
attrs = ['' for c in columns]
for row in rows:
if float(row[2]) < 0.:
faster_queries += 1
attrs[2] = 'style="background: #adbdff"'
else:
slower_queries += 1
attrs[2] = 'style="background: #ffb0a0"'
print(tableRow(row, attrs))
print(tableEnd())
print_changes()
slow_on_client_rows = tsvRows('slow-on-client.tsv')
error_tests += len(slow_on_client_rows)
printSimpleTable('Slow on client',
['Client time, s', 'Server time, s', 'Ratio', 'Query'],
slow_on_client_rows)
def print_unstable_queries():
global unstable_queries
global very_unstable_queries
unstable_rows = tsvRows('unstable-queries.tsv')
if not unstable_rows:
return
unstable_queries += len(unstable_rows)
columns = [
'Old, s', #0
'New, s', #1
'Relative difference (new&nbsp;-&nbsp;old)/old', #2
'Randomization distribution quantiles [5%,&nbsp;50%,&nbsp;95%,&nbsp;99%]', #3
'Test', #4
'Query' #5
]
print(tableStart('Unstable queries'))
print(tableHeader(columns))
attrs = ['' for c in columns]
for r in unstable_rows:
rd = ast.literal_eval(r[3])
# Note the zero-based array index, this is rd[3] in SQL.
if rd[2] > 0.2:
very_unstable_queries += 1
attrs[3] = 'style="background: #ffb0a0"'
else:
attrs[3] = ''
print(tableRow(r, attrs))
print(tableEnd())
print_unstable_queries()
run_error_rows = tsvRows('run-errors.tsv')
error_tests += len(run_error_rows)
printSimpleTable('Run errors', ['Test', 'Error'], run_error_rows)
skipped_tests_rows = tsvRows('skipped-tests.tsv')
printSimpleTable('Skipped tests', ['Test', 'Reason'], skipped_tests_rows)
printSimpleTable('Tests with most unstable queries',
['Test', 'Unstable', 'Changed perf', 'Total not OK'],
tsvRows('bad-tests.tsv'))
def print_test_times():
global slow_average_tests
rows = tsvRows('test-times.tsv')
if not rows:
return
columns = [
'Test', #0
'Wall clock time, s', #1
'Total client time, s', #2
'Total queries', #3
'Ignored short queries', #4
'Longest query<br>(sum for all runs), s', #5
'Avg wall clock time<br>(sum for all runs), s', #6
'Shortest query<br>(sum for all runs), s', #7
]
print(tableStart('Test times'))
print(tableHeader(columns))
attrs = ['' for c in columns]
for r in rows:
if float(r[6]) > 22:
# FIXME should be 15s max -- investigate parallel_insert
slow_average_tests += 1
attrs[6] = 'style="background: #ffb0a0"'
else:
attrs[6] = ''
if float(r[5]) > 30:
slow_average_tests += 1
attrs[5] = 'style="background: #ffb0a0"'
else:
attrs[5] = ''
print(tableRow(r, attrs))
print(tableEnd())
print_test_times()
# Add the errors reported by various steps of comparison script
report_errors += [l.strip() for l in open('report-errors.rep')]
if len(report_errors):
print(tableStart('Errors while building the report'))
print(tableHeader(['Error']))
for x in report_errors:
print(tableRow([x]))
print(tableEnd())
print("""
<p class="links">
<a href="output.7z">Test output</a>
<a href="compare.log">Log</a>
</p>
</body>
</html>
""")
status = 'success'
message = 'See the report'
message_array = []
if slow_average_tests:
status = 'failure'
message_array.append(str(slow_average_tests) + ' too long')
if faster_queries:
message_array.append(str(faster_queries) + ' faster')
if slower_queries:
status = 'failure'
message_array.append(str(slower_queries) + ' slower')
if unstable_queries:
message_array.append(str(unstable_queries) + ' unstable')
if very_unstable_queries:
status = 'failure'
error_tests += slow_average_tests
if error_tests:
status = 'failure'
message_array.append(str(error_tests) + ' errors')
if message_array:
message = ', '.join(message_array)
if report_errors:
status = 'failure'
message = 'Errors while building the report.'
print("""
<!--status: {status}-->
<!--message: {message}-->
""".format(status=status, message=message))
print(tableStart('Unstable queries'))
print(tableHeader(columns))
attrs = ['' for c in columns]
for r in unstable_rows:
rd = ast.literal_eval(r[3])
# Note the zero-based array index, this is rd[3] in SQL.
if rd[2] > 0.2:
very_unstable_queries += 1
attrs[3] = 'style="background: #ffb0a0"'
else:
attrs[3] = ''
print(tableRow(r, attrs))
print(tableEnd())
print_unstable_queries()
run_error_rows = tsvRows('run-errors.tsv')
error_tests += len(run_error_rows)
printSimpleTable('Run errors', ['Test', 'Error'], run_error_rows)
skipped_tests_rows = tsvRows('skipped-tests.tsv')
printSimpleTable('Skipped tests', ['Test', 'Reason'], skipped_tests_rows)
printSimpleTable('Tests with most unstable queries',
['Test', 'Unstable', 'Changed perf', 'Total not OK'],
tsvRows('bad-tests.tsv'))
def print_test_times():
global slow_average_tests
rows = tsvRows('test-times.tsv')
if not rows:
return
columns = [
'Test', #0
'Wall clock time, s', #1
'Total client time, s', #2
'Total queries', #3
'Ignored short queries', #4
'Longest query<br>(sum for all runs), s', #5
'Avg wall clock time<br>(sum for all runs), s', #6
'Shortest query<br>(sum for all runs), s', #7
]
print(tableStart('Test times'))
print(tableHeader(columns))
attrs = ['' for c in columns]
for r in rows:
if float(r[6]) > 22:
# FIXME should be 15s max -- investigate parallel_insert
slow_average_tests += 1
attrs[6] = 'style="background: #ffb0a0"'
else:
attrs[6] = ''
if float(r[5]) > 30:
slow_average_tests += 1
attrs[5] = 'style="background: #ffb0a0"'
else:
attrs[5] = ''
print(tableRow(r, attrs))
print(tableEnd())
print_test_times()
# Add the errors reported by various steps of comparison script
report_errors += [l.strip() for l in open('report-errors.rep')]
if len(report_errors):
print(tableStart('Errors while building the report'))
print(tableHeader(['Error']))
for x in report_errors:
print(tableRow([x]))
print(tableEnd())
print("""
<p class="links">
<a href="output.7z">Test output</a>
<a href="all-queries.html">All queries</a>
<a href="compare.log">Log</a>
</p>
</body>
</html>
""")
status = 'success'
message = 'See the report'
message_array = []
if slow_average_tests:
status = 'failure'
message_array.append(str(slow_average_tests) + ' too long')
if faster_queries:
message_array.append(str(faster_queries) + ' faster')
if slower_queries:
status = 'failure'
message_array.append(str(slower_queries) + ' slower')
if unstable_queries:
message_array.append(str(unstable_queries) + ' unstable')
if very_unstable_queries:
status = 'failure'
error_tests += slow_average_tests
if error_tests:
status = 'failure'
message_array.append(str(error_tests) + ' errors')
if message_array:
message = ', '.join(message_array)
if report_errors:
status = 'failure'
message = 'Errors while building the report.'
print("""
<!--status: {status}-->
<!--message: {message}-->
""".format(status=status, message=message))
elif args.report == 'all-queries':
print(header_template.format())
printSimpleTable('Tested commits', ['Old', 'New'],
[['<pre>{}</pre>'.format(x) for x in
[open('left-commit.txt').read(),
open('right-commit.txt').read()]]])
def print_all_queries():
rows = tsvRows('all-queries.tsv')
if not rows:
return
columns = [
'Old, s', #0
'New, s', #1
'Relative difference (new&nbsp;-&nbsp;old)/old', #2
'Times speedup/slowdown', #3
'Randomization distribution quantiles \
[5%,&nbsp;50%,&nbsp;95%,&nbsp;99%]', #4
'Test', #5
'Query', #6
]
print(tableStart('All query times'))
print(tableHeader(columns))
attrs = ['' for c in columns]
for r in rows:
if float(r[2]) > 0.05:
attrs[3] = 'style="background: #ffb0a0"'
elif float(r[2]) < -0.05:
attrs[3] = 'style="background: #adbdff"'
else:
attrs[3] = ''
print(tableRow(r, attrs))
print(tableEnd())
print_all_queries()
print("""
<p class="links">
<a href="output.7z">Test output</a>
<a href="report.html">Main report</a>
<a href="compare.log">Log</a>
</p>
</body>
</html>
""")
tests/performance/array_fill.xml
浏览文件 @ d4445485
<test>
<stop_conditions>
<all_of>
<total_time_ms>10000</total_time_ms>
</all_of>
</stop_conditions>
<query>SELECT arraySlice(arrayFill(x -> ((x % 2) >= 0), range(100000000)), 1, 10)</query>
<query>SELECT arraySlice(arrayFill(x -> (((x.1) % 2) >= 0), arrayMap(x -> (x, toString(x)), range(100000000))), 1, 10)</query>
<query>SELECT arraySlice(arrayFill(x -> ((x % 2) >= 2), range(100000000)), 1, 10)</query>
<query>SELECT arraySlice(arrayFill(x -> (((x.1) % 2) >= 2), arrayMap(x -> (x, toString(x)), range(100000000))), 1, 10)</query>
<query>SELECT arraySlice(arrayFill(x -> ((x % 2) = 0), range(100000000)), 1, 10)</query>
<query>SELECT arraySlice(arrayFill(x -> (((x.1) % 2) = 0), arrayMap(x -> (x, toString(x)), range(100000000))), 1, 10)</query>
<query>SELECT arraySlice(arrayFill(x -> ((x % 2) >= 0), range(100000000)), 1, 10) FORMAT Null</query>
<query>SELECT arraySlice(arrayFill(x -> (((x.1) % 2) >= 0), arrayMap(x -> (x, toString(x)), range(100000000))), 1, 10) FORMAT Null</query>
<query>SELECT arraySlice(arrayFill(x -> ((x % 2) >= 2), range(100000000)), 1, 10) FORMAT Null</query>
<query>SELECT arraySlice(arrayFill(x -> (((x.1) % 2) >= 2), arrayMap(x -> (x, toString(x)), range(100000000))), 1, 10) FORMAT Null</query>
<query>SELECT arraySlice(arrayFill(x -> ((x % 2) = 0), range(100000000)), 1, 10) FORMAT Null</query>
<query>SELECT arraySlice(arrayFill(x -> (((x.1) % 2) = 0), arrayMap(x -> (x, toString(x)), range(100000000))), 1, 10) FORMAT Null</query>
</test>
tests/performance/concat_hits.xml
浏览文件 @ d4445485
<test>
<stop_conditions>
<all_of>
<iterations>5</iterations>
<min_time_not_changing_for_ms>10000</min_time_not_changing_for_ms>
</all_of>
<any_of>
<iterations>50</iterations>
<total_time_ms>60000</total_time_ms>
</any_of>
</stop_conditions>
<preconditions>
<table_exists>test.hits</table_exists>
</preconditions>
......@@ -24,13 +11,13 @@
<query>SELECT count() FROM test.hits WHERE NOT ignore(concat(MobilePhoneModel, 'Hello'))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(concat(PageCharset, 'a'))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}{}', URL, URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}{}', URL, SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}{}', MobilePhoneModel, SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}Hello', URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('World{}', SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}Hello', MobilePhoneModel))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}a', PageCharset))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}{{}}', URL, URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}{{}}', URL, SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}{{}}', MobilePhoneModel, SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}Hello', URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('World{{}}', SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}Hello', MobilePhoneModel))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}a', PageCharset))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(concat(URL, URL, URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(concat(URL, SearchPhrase, MobilePhoneModel))</query>
......@@ -39,10 +26,10 @@
<query>SELECT count() FROM test.hits WHERE NOT ignore(concat(MobilePhoneModel, 'Hello', PageCharset))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(concat('a', PageCharset, 'b'))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}{}{}', URL, URL, URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}{}{}', URL, SearchPhrase, MobilePhoneModel))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}Hello{}', URL, URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('Hello{}World', SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{}Hello{}', MobilePhoneModel, PageCharset))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('a{}b', PageCharset))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}{{}}{{}}', URL, URL, URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}{{}}{{}}', URL, SearchPhrase, MobilePhoneModel))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}Hello{{}}', URL, URL))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('Hello{{}}World', SearchPhrase))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('{{}}Hello{{}}', MobilePhoneModel, PageCharset))</query>
<query>SELECT count() FROM test.hits WHERE NOT ignore(format('a{{}}b', PageCharset))</query>
</test>
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或