Add ramdrive

This commit is contained in:
alesapin 2021-12-09 12:04:05 +03:00
parent b62de9b2f4
commit 67d89929a3
3 changed files with 87 additions and 84 deletions

View File

@ -89,6 +89,8 @@ jobs:
REPORTS_PATH: ${{runner.temp}}/reports_dir
CHECK_NAME: 'Performance Comparison (actions)'
REPO_COPY: ${{runner.temp}}/performance_comparison/ClickHouse
RAMDRIVE_PATH: ${{runner.temp}}/performance_comparison/ramdrive
RAMDRIVE_SIZE: '120G'
run: |
sudo rm -fr $TEMP_PATH
mkdir -p $TEMP_PATH

View File

@ -92,3 +92,6 @@ def download_unit_tests(check_name, reports_path, result_path):
def download_clickhouse_binary(check_name, reports_path, result_path):
download_builds_filter(check_name, reports_path, result_path, lambda x: x.endswith('clickhouse'))
def download_performance_build(check_name, reports_path, result_path):
download_builds_filter(check_name, reports_path, result_path, lambda x: x.endswith('performance.tgz'))

View File

@ -13,11 +13,9 @@ from github import Github
from pr_info import PRInfo
from s3_helper import S3Helper
from ci_config import build_config_to_string
from get_robot_token import get_best_robot_token
from docker_pull_helper import get_image_with_version
from commit_status_helper import get_commit, post_commit_status
from build_download_helper import get_build_config_for_check, get_build_urls
IMAGE_NAME = 'clickhouse/performance-comparison'
@ -27,10 +25,27 @@ def get_run_command(workspace, result_path, pr_to_test, sha_to_test, additional_
f"-e PR_TO_TEST={pr_to_test} -e SHA_TO_TEST={sha_to_test} {additional_env} " \
f"{image}"
class RamDrive:
def __init__(self, path, size):
self.path = path
self.size = size
def __enter__(self):
if not os.path.exists(self.path):
os.makedirs(self.path)
subprocess.check_call(f"sudo mount -t tmpfs -o rw,size={self.size} tmpfs {self.path}", shell=True)
def __exit__(self, exc_type, exc_val, exc_tb):
subprocess.check_call(f"sudo unmount {self.path}", shell=True)
subprocess.check_call(f"sudo rm -fr {self.path}", shell=True)
if __name__ == "__main__":
logging.basicConfig(level=logging.INFO)
temp_path = os.getenv("TEMP_PATH", os.path.abspath("."))
repo_path = os.getenv("REPO_COPY", os.path.abspath("../../"))
ramdrive_path = os.getenv("RAMDRIVE_PATH", os.path.join(temp_path, "ramdrive"))
ramdrive_size = os.getenv("RAMDRIVE_SIZE", '120G')
reports_path = os.getenv("REPORTS_PATH", "./reports")
check_name = sys.argv[1]
@ -45,21 +60,6 @@ if __name__ == "__main__":
pr_info = PRInfo(event)
commit = get_commit(gh, pr_info.sha)
build_config = get_build_config_for_check(check_name)
print(build_config)
build_config_str = build_config_to_string(build_config)
print(build_config_str)
urls = get_build_urls(build_config_str, reports_path)
if not urls:
raise Exception("No build URLs found")
for url in urls:
if url.endswith('/performance.tgz'):
build_url = url
break
else:
raise Exception("Cannot binary clickhouse among build results")
docker_env = ''
docker_env += " -e S3_URL=https://s3.amazonaws.com/clickhouse-builds"
@ -75,84 +75,82 @@ if __name__ == "__main__":
docker_image = get_image_with_version(reports_path, IMAGE_NAME)
result_path = os.path.join(temp_path, 'result_path')
if not os.path.exists(result_path):
os.makedirs(result_path)
with RamDrive(ramdrive_path, ramdrive_size):
result_path = ramdrive_path
run_command = get_run_command(result_path, result_path, pr_info.number, pr_info.sha, docker_env, docker_image)
logging.info("Going to run command %s", run_command)
run_log_path = os.path.join(temp_path, "runlog.log")
with open(run_log_path, 'w', encoding='utf-8') as log:
with subprocess.Popen(run_command, shell=True, stderr=log, stdout=log) as process:
retcode = process.wait()
if retcode == 0:
logging.info("Run successfully")
else:
logging.info("Run failed")
run_command = get_run_command(result_path, result_path, pr_info.number, pr_info.sha, docker_env, docker_image)
logging.info("Going to run command %s", run_command)
run_log_path = os.path.join(temp_path, "runlog.log")
with open(run_log_path, 'w', encoding='utf-8') as log:
with subprocess.Popen(run_command, shell=True, stderr=log, stdout=log) as process:
retcode = process.wait()
if retcode == 0:
logging.info("Run successfully")
else:
logging.info("Run failed")
subprocess.check_call(f"sudo chown -R ubuntu:ubuntu {temp_path}", shell=True)
subprocess.check_call(f"sudo chown -R ubuntu:ubuntu {temp_path}", shell=True)
s3_prefix = f'{pr_info.number}/{pr_info.sha}/performance_comparison/'
paths = {
'compare.log': 'compare.log',
'output.7z': 'output.7z',
'report.html': 'report.html',
'all-queries.html': 'all-queries.html',
'queries.rep': 'queries.rep',
'all-query-metrics.tsv': 'report/all-query-metrics.tsv',
}
s3_prefix = f'{pr_info.number}/{pr_info.sha}/performance_comparison/'
paths = {
'compare.log': 'compare.log',
'output.7z': 'output.7z',
'report.html': 'report.html',
'all-queries.html': 'all-queries.html',
'queries.rep': 'queries.rep',
'all-query-metrics.tsv': 'report/all-query-metrics.tsv',
}
s3_helper = S3Helper('https://s3.amazonaws.com')
for file in paths:
try:
paths[file] = s3_helper.upload_test_report_to_s3(
os.path.join(result_path, paths[file]),
s3_prefix + file)
except Exception:
paths[file] = ''
traceback.print_exc()
s3_helper = S3Helper('https://s3.amazonaws.com')
for file in paths:
# Upload all images and flamegraphs to S3
try:
paths[file] = s3_helper.upload_test_report_to_s3(
os.path.join(result_path, paths[file]),
s3_prefix + file)
s3_helper.upload_test_folder_to_s3(
os.path.join(result_path, 'images'),
s3_prefix + 'images'
)
except Exception:
paths[file] = ''
traceback.print_exc()
# Upload all images and flamegraphs to S3
try:
s3_helper.upload_test_folder_to_s3(
os.path.join(result_path, 'images'),
s3_prefix + 'images'
)
except Exception:
traceback.print_exc()
# Try to fetch status from the report.
status = ''
message = ''
try:
report_text = open(os.path.join(result_path, 'report.html'), 'r').read()
status_match = re.search('<!--[ ]*status:(.*)-->', report_text)
message_match = re.search('<!--[ ]*message:(.*)-->', report_text)
if status_match:
status = status_match.group(1).strip()
if message_match:
message = message_match.group(1).strip()
except Exception:
traceback.print_exc()
status = 'failure'
message = 'Failed to parse the report.'
# Try to fetch status from the report.
status = ''
message = ''
try:
report_text = open(os.path.join(result_path, 'report.html'), 'r').read()
status_match = re.search('<!--[ ]*status:(.*)-->', report_text)
message_match = re.search('<!--[ ]*message:(.*)-->', report_text)
if status_match:
status = status_match.group(1).strip()
if message_match:
message = message_match.group(1).strip()
except Exception:
traceback.print_exc()
status = 'failure'
message = 'Failed to parse the report.'
if not status:
status = 'failure'
message = 'No status in report.'
elif not message:
status = 'failure'
message = 'No message in report.'
if not status:
status = 'failure'
message = 'No status in report.'
elif not message:
status = 'failure'
message = 'No message in report.'
report_url = task_url
report_url = task_url
if paths['compare.log']:
report_url = paths['compare.log']
if paths['compare.log']:
report_url = paths['compare.log']
if paths['output.7z']:
report_url = paths['output.7z']
if paths['output.7z']:
report_url = paths['output.7z']
if paths['report.html']:
report_url = paths['report.html']
if paths['report.html']:
report_url = paths['report.html']
post_commit_status(gh, pr_info.sha, check_name, message, status, report_url)
post_commit_status(gh, pr_info.sha, check_name, message, status, report_url)