ClickHouse/tests/integration/ci-runner.py
2021-03-23 14:35:45 +03:00

592 lines
25 KiB
Python
Executable File

#!/usr/bin/env python3
import logging
import subprocess
import os
import time
import shutil
from collections import defaultdict
import random
import json
import csv
MAX_RETRY = 2
SLEEP_BETWEEN_RETRIES = 5
CLICKHOUSE_BINARY_PATH = "/usr/bin/clickhouse"
CLICKHOUSE_ODBC_BRIDGE_BINARY_PATH = "/usr/bin/clickhouse-odbc-bridge"
DOCKERD_LOGS_PATH = "/ClickHouse/tests/integration/dockerd.log"
TRIES_COUNT = 10
MAX_TIME_SECONDS = 3600
MAX_TIME_IN_SANDBOX = 20 * 60 # 20 minutes
TASK_TIMEOUT = 8 * 60 * 60 # 8 hours
def get_tests_to_run(pr_info):
result = set([])
changed_files = pr_info['changed_files']
if changed_files is None:
return []
for fpath in changed_files:
if 'tests/integration/test_' in fpath:
logging.info('File %s changed and seems like integration test', fpath)
result.add(fpath.split('/')[2])
return list(result)
def filter_existing_tests(tests_to_run, repo_path):
result = []
for relative_test_path in tests_to_run:
if os.path.exists(os.path.join(repo_path, 'tests/integration', relative_test_path)):
result.append(relative_test_path)
else:
logging.info("Skipping test %s, seems like it was removed", relative_test_path)
return result
def _get_deselect_option(tests):
return ' '.join(['--deselect {}'.format(t) for t in tests])
def parse_test_results_output(fname):
read = False
description_output = []
with open(fname, 'r') as out:
for line in out:
if read and line.strip() and not line.startswith('=='):
description_output.append(line.strip())
if 'short test summary info' in line:
read = True
return description_output
def get_counters(output):
counters = {
"ERROR": set([]),
"PASSED": set([]),
"FAILED": set([]),
}
for line in output:
if '.py' in line:
line_arr = line.strip().split(' ')
state = line_arr[0]
test_name = ' '.join(line_arr[1:])
if ' - ' in test_name:
test_name = test_name[:test_name.find(' - ')]
if state in counters:
counters[state].add(test_name)
else:
logging.info("Strange line %s", line)
else:
logging.info("Strange line %s")
return {k: list(v) for k, v in counters.items()}
def parse_test_times(fname):
read = False
description_output = []
with open(fname, 'r') as out:
for line in out:
if read and '==' in line:
break
if read and line.strip():
description_output.append(line.strip())
if 'slowest durations' in line:
read = True
return description_output
def get_test_times(output):
result = defaultdict(float)
for line in output:
if '.py' in line:
line_arr = line.strip().split(' ')
test_time = line_arr[0]
test_name = ' '.join([elem for elem in line_arr[2:] if elem])
if test_name not in result:
result[test_name] = 0.0
result[test_name] += float(test_time[:-1])
return result
def clear_ip_tables_and_restart_daemons():
logging.info("Dump iptables after run %s", subprocess.check_output("iptables -L", shell=True))
try:
logging.info("Killing all alive docker containers")
subprocess.check_output("docker kill $(docker ps -q)", shell=True)
except subprocess.CalledProcessError as err:
logging.info("docker kill excepted: " + str(err))
try:
logging.info("Removing all docker containers")
subprocess.check_output("docker rm $(docker ps -a -q) --force", shell=True)
except subprocess.CalledProcessError as err:
logging.info("docker rm excepted: " + str(err))
try:
logging.info("Stopping docker daemon")
subprocess.check_output("service docker stop", shell=True)
except subprocess.CalledProcessError as err:
logging.info("docker stop excepted: " + str(err))
try:
for i in range(200):
try:
logging.info("Restarting docker %s", i)
subprocess.check_output("service docker start", shell=True)
subprocess.check_output("docker ps", shell=True)
break
except subprocess.CalledProcessError as err:
time.sleep(0.5)
logging.info("Waiting docker to start, current %s", str(err))
else:
raise Exception("Docker daemon doesn't responding")
except subprocess.CalledProcessError as err:
logging.info("Can't reload docker: " + str(err))
iptables_iter = 0
try:
for i in range(1000):
iptables_iter = i
# when rules will be empty, it will raise exception
subprocess.check_output("iptables -D DOCKER-USER 1", shell=True)
except subprocess.CalledProcessError as err:
logging.info("All iptables rules cleared, " + str(iptables_iter) + "iterations, last error: " + str(err))
class ClickhouseIntegrationTestsRunner:
def __init__(self, result_path, params):
self.result_path = result_path
self.params = params
self.image_versions = self.params['docker_images_with_versions']
self.shuffle_groups = self.params['shuffle_test_groups']
self.flaky_check = 'flaky check' in self.params['context_name']
self.start_time = time.time()
self.soft_deadline_time = self.start_time + (TASK_TIMEOUT - MAX_TIME_IN_SANDBOX)
def path(self):
return self.result_path
def base_path(self):
return os.path.join(str(self.result_path), '../')
def should_skip_tests(self):
return []
def get_image_with_version(self, name):
if name in self.image_versions:
return name + ":" + self.image_versions[name]
logging.warn("Cannot find image %s in params list %s", name, self.image_versions)
if ':' not in name:
return name + ":latest"
return name
def get_single_image_version(self):
name = self.get_images_names()[0]
if name in self.image_versions:
return self.image_versions[name]
logging.warn("Cannot find image %s in params list %s", name, self.image_versions)
return 'latest'
def shuffle_test_groups(self):
return self.shuffle_groups != 0
@staticmethod
def get_images_names():
return ["yandex/clickhouse-integration-tests-runner", "yandex/clickhouse-mysql-golang-client",
"yandex/clickhouse-mysql-java-client", "yandex/clickhouse-mysql-js-client",
"yandex/clickhouse-mysql-php-client", "yandex/clickhouse-postgresql-java-client",
"yandex/clickhouse-integration-test", "yandex/clickhouse-kerberos-kdc",
"yandex/clickhouse-integration-helper", ]
def _can_run_with(self, path, opt):
with open(path, 'r') as script:
for line in script:
if opt in line:
return True
return False
def _install_clickhouse(self, debs_path):
for package in ('clickhouse-common-static_', 'clickhouse-server_', 'clickhouse-client', 'clickhouse-common-static-dbg_'): # order matters
logging.info("Installing package %s", package)
for f in os.listdir(debs_path):
if package in f:
full_path = os.path.join(debs_path, f)
logging.info("Package found in %s", full_path)
log_name = "install_" + f + ".log"
log_path = os.path.join(str(self.path()), log_name)
with open(log_path, 'w') as log:
cmd = "dpkg -i {}".format(full_path)
logging.info("Executing installation cmd %s", cmd)
retcode = subprocess.Popen(cmd, shell=True, stderr=log, stdout=log).wait()
if retcode == 0:
logging.info("Instsallation of %s successfull", full_path)
else:
raise Exception("Installation of %s failed", full_path)
break
else:
raise Exception("Package with {} not found".format(package))
logging.info("Unstripping binary")
# logging.info("Unstring %s", subprocess.check_output("eu-unstrip /usr/bin/clickhouse {}".format(CLICKHOUSE_BINARY_PATH), shell=True))
logging.info("All packages installed")
os.chmod(CLICKHOUSE_BINARY_PATH, 0o777)
os.chmod(CLICKHOUSE_ODBC_BRIDGE_BINARY_PATH, 0o777)
result_path_bin = os.path.join(str(self.base_path()), "clickhouse")
result_path_bridge = os.path.join(str(self.base_path()), "clickhouse-odbc-bridge")
shutil.copy(CLICKHOUSE_BINARY_PATH, result_path_bin)
shutil.copy(CLICKHOUSE_ODBC_BRIDGE_BINARY_PATH, result_path_bridge)
return None, None
def _compress_logs(self, path, result_path):
subprocess.check_call("tar czf {} -C {} .".format(result_path, path), shell=True) # STYLE_CHECK_ALLOW_SUBPROCESS_CHECK_CALL
def _get_all_tests(self, repo_path):
image_cmd = self._get_runner_image_cmd(repo_path)
cmd = "cd {}/tests/integration && ./runner {} ' --setup-plan' | grep '::' | sed 's/ (fixtures used:.*//g' | sed 's/^ *//g' > all_tests.txt".format(repo_path, image_cmd)
logging.info("Getting all tests with cmd '%s'", cmd)
subprocess.check_call(cmd, shell=True) # STYLE_CHECK_ALLOW_SUBPROCESS_CHECK_CALL
all_tests_file_path = "{}/tests/integration/all_tests.txt".format(repo_path)
if not os.path.isfile(all_tests_file_path) or os.path.getsize(all_tests_file_path) == 0:
raise Exception("There is something wrong with getting all tests list: file '{}' is empty or does not exist.".format(all_tests_file_path))
all_tests = []
with open(all_tests_file_path, "r") as all_tests_file:
for line in all_tests_file:
all_tests.append(line.strip())
return list(sorted(all_tests))
def _get_parallel_tests(self, repo_path):
parallel_tests_file_path = "{}/tests/integration/parallel.txt".format(repo_path)
if not os.path.isfile(parallel_tests_file_path) or os.path.getsize(parallel_tests_file_path) == 0:
raise Exception("There is something wrong with getting all tests list: file '{}' is empty or does not exist.".format(parallel_tests_file_path))
parallel_tests = []
with open(parallel_tests_file_path, "r") as parallel_tests_file:
for line in parallel_tests_file:
parallel_tests.append(line.strip())
return list(sorted(parallel_tests))
def group_test_by_file(self, tests):
result = {}
for test in tests:
test_file = test.split('::')[0]
if test_file not in result:
result[test_file] = []
result[test_file].append(test)
return result
def _update_counters(self, main_counters, current_counters):
for test in current_counters["PASSED"]:
if test not in main_counters["PASSED"] and test not in main_counters["FLAKY"]:
is_flaky = False
if test in main_counters["FAILED"]:
main_counters["FAILED"].remove(test)
is_flaky = True
if test in main_counters["ERROR"]:
main_counters["ERROR"].remove(test)
is_flaky = True
if is_flaky:
main_counters["FLAKY"].append(test)
else:
main_counters["PASSED"].append(test)
for state in ("ERROR", "FAILED"):
for test in current_counters[state]:
if test in main_counters["FLAKY"]:
continue
if test in main_counters["PASSED"]:
main_counters["PASSED"].remove(test)
main_counters["FLAKY"].append(test)
continue
if test not in main_counters[state]:
main_counters[state].append(test)
def _get_runner_image_cmd(self, repo_path):
image_cmd = ''
if self._can_run_with(os.path.join(repo_path, "tests/integration", "runner"), '--docker-image-version'):
for img in self.get_images_names():
if img == "yandex/clickhouse-integration-tests-runner":
runner_version = self.get_single_image_version()
logging.info("Can run with custom docker image version %s", runner_version)
image_cmd += ' --docker-image-version={} '.format(runner_version)
else:
if self._can_run_with(os.path.join(repo_path, "tests/integration", "runner"), '--docker-compose-images-tags'):
image_cmd += '--docker-compose-images-tags={} '.format(self.get_image_with_version(img))
else:
image_cmd = ''
logging.info("Cannot run with custom docker image version :(")
return image_cmd
def run_test_group(self, repo_path, test_group, tests_in_group, num_tries):
counters = {
"ERROR": [],
"PASSED": [],
"FAILED": [],
"SKIPPED": [],
"FLAKY": [],
}
tests_times = defaultdict(float)
if self.soft_deadline_time < time.time():
for test in tests_in_group:
logging.info("Task timeout exceeded, skipping %s", test)
counters["SKIPPED"].append(test)
tests_times[test] = 0
log_name = None
log_path = None
return counters, tests_times, log_name, log_path
image_cmd = self._get_runner_image_cmd(repo_path)
test_group_str = test_group.replace('/', '_').replace('.', '_')
for i in range(num_tries):
logging.info("Running test group %s for the %s retry", test_group, i)
clear_ip_tables_and_restart_daemons()
output_path = os.path.join(str(self.path()), "test_output_" + test_group_str + "_" + str(i) + ".log")
log_name = "integration_run_" + test_group_str + "_" + str(i) + ".txt"
log_path = os.path.join(str(self.path()), log_name)
logging.info("Will wait output inside %s", output_path)
test_names = set([])
for test_name in tests_in_group:
if test_name not in counters["PASSED"]:
if '[' in test_name:
test_names.add(test_name[:test_name.find('[')])
else:
test_names.add(test_name)
test_cmd = ' '.join([test for test in sorted(test_names)])
cmd = "cd {}/tests/integration && ./runner {} -t {} --parallel 10 '-ss -rfEp --color=no --durations=0 {}' | tee {}".format(
repo_path, image_cmd, test_cmd, _get_deselect_option(self.should_skip_tests()), output_path)
with open(log_path, 'w') as log:
logging.info("Executing cmd: %s", cmd)
retcode = subprocess.Popen(cmd, shell=True, stderr=log, stdout=log).wait()
if retcode == 0:
logging.info("Run %s group successfully", test_group)
else:
logging.info("Some tests failed")
if os.path.exists(output_path):
lines = parse_test_results_output(output_path)
new_counters = get_counters(lines)
times_lines = parse_test_times(output_path)
new_tests_times = get_test_times(times_lines)
self._update_counters(counters, new_counters)
for test_name, test_time in new_tests_times.items():
tests_times[test_name] = test_time
os.remove(output_path)
if len(counters["PASSED"]) + len(counters["FLAKY"]) == len(tests_in_group):
logging.info("All tests from group %s passed", test_group)
break
if len(counters["PASSED"]) + len(counters["FLAKY"]) >= 0 and len(counters["FAILED"]) == 0 and len(counters["ERROR"]) == 0:
logging.info("Seems like all tests passed but some of them are skipped or deselected. Ignoring them and finishing group.")
break
else:
for test in tests_in_group:
if test not in counters["PASSED"] and test not in counters["ERROR"] and test not in counters["FAILED"]:
counters["ERROR"].append(test)
return counters, tests_times, log_name, log_path
def run_flaky_check(self, repo_path, build_path):
pr_info = self.params['pr_info']
# pytest swears, if we require to run some tests which was renamed or deleted
tests_to_run = filter_existing_tests(get_tests_to_run(pr_info), repo_path)
if not tests_to_run:
logging.info("No tests to run found")
return 'success', 'Nothing to run', [('Nothing to run', 'OK')], ''
self._install_clickhouse(build_path)
logging.info("Found '%s' tests to run", ' '.join(tests_to_run))
result_state = "success"
description_prefix = "No flaky tests: "
start = time.time()
logging.info("Starting check with retries")
final_retry = 0
log_paths = []
for i in range(TRIES_COUNT):
final_retry += 1
logging.info("Running tests for the %s time", i)
counters, tests_times, _, log_path = self.run_test_group(repo_path, "flaky", tests_to_run, 1)
log_paths.append(log_path)
if counters["FAILED"]:
logging.info("Found failed tests: %s", ' '.join(counters["FAILED"]))
description_prefix = "Flaky tests found: "
result_state = "failure"
break
if counters["ERROR"]:
description_prefix = "Flaky tests found: "
logging.info("Found error tests: %s", ' '.join(counters["ERROR"]))
# NOTE "error" result state will restart the whole test task, so we use "failure" here
result_state = "failure"
break
assert len(counters["FLAKY"]) == 0
logging.info("Try is OK, all tests passed, going to clear env")
clear_ip_tables_and_restart_daemons()
logging.info("And going to sleep for some time")
if time.time() - start > MAX_TIME_SECONDS:
logging.info("Timeout reached, going to finish flaky check")
break
time.sleep(5)
logging.info("Finally all tests done, going to compress test dir")
test_logs = os.path.join(str(self.path()), "./test_dir.tar")
self._compress_logs("{}/tests/integration".format(repo_path), test_logs)
logging.info("Compression finished")
result_path_dockerd_logs = os.path.join(str(self.path()), "dockerd.log")
if os.path.exists(result_path_dockerd_logs):
shutil.copy(DOCKERD_LOGS_PATH, result_path_dockerd_logs)
test_result = []
for state in ("ERROR", "FAILED", "PASSED", "SKIPPED", "FLAKY"):
if state == "PASSED":
text_state = "OK"
elif state == "FAILED":
text_state = "FAIL"
else:
text_state = state
test_result += [(c + ' (✕' + str(final_retry) + ')', text_state, "{:.2f}".format(tests_times[c])) for c in counters[state]]
status_text = description_prefix + ', '.join([str(n).lower().replace('failed', 'fail') + ': ' + str(len(c)) for n, c in counters.items()])
return result_state, status_text, test_result, [test_logs] + log_paths
def run_impl(self, repo_path, build_path):
if self.flaky_check:
return self.run_flaky_check(repo_path, build_path)
self._install_clickhouse(build_path)
logging.info("Dump iptables before run %s", subprocess.check_output("iptables -L", shell=True))
all_tests = self._get_all_tests(repo_path)
parallel_tests = self._get_parallel_tests(repo_path)
logging.info("Found %s tests first 3 %s", len(all_tests), ' '.join(all_tests[:3]))
filtered_parallel_tests = list(filter(lambda test: test in all_tests, parallel_tests))
filtered_unparallel_tests = list(filter(lambda test: test not in parallel_tests, all_tests))
not_found_tests = list(filter(lambda test: test not in all_tests, parallel_tests))
logging.info("Found %s tests first 3 %s, parallel %s, other %s", len(all_tests), ' '.join(all_tests[:3]), len(filtered_parallel_tests), len(filtered_unparallel_tests))
logging.info("Not found %s tests first 3 %s", len(not_found_tests), ' '.join(not_found_tests[:3]))
grouped_tests = self.group_test_by_file(filtered_unparallel_tests)
grouped_tests["parallel"] = filtered_parallel_tests
logging.info("Found %s tests groups", len(grouped_tests))
counters = {
"ERROR": [],
"PASSED": [],
"FAILED": [],
"SKIPPED": [],
"FLAKY": [],
}
tests_times = defaultdict(float)
logs = []
items_to_run = list(grouped_tests.items())
logging.info("Total test groups %s", len(items_to_run))
if self.shuffle_test_groups():
logging.info("Shuffling test groups")
random.shuffle(items_to_run)
for group, tests in items_to_run:
logging.info("Running test group %s countaining %s tests", group, len(tests))
group_counters, group_test_times, _, log_path = self.run_test_group(repo_path, group, tests, MAX_RETRY)
total_tests = 0
for counter, value in group_counters.items():
logging.info("Tests from group %s stats, %s count %s", group, counter, len(value))
counters[counter] += value
logging.info("Totally have %s with status %s", len(counters[counter]), counter)
total_tests += len(counters[counter])
logging.info("Totally finished tests %s/%s", total_tests, len(all_tests))
for test_name, test_time in group_test_times.items():
tests_times[test_name] = test_time
logs.append(log_path)
if len(counters["FAILED"]) + len(counters["ERROR"]) >= 20:
logging.info("Collected more than 20 failed/error tests, stopping")
break
logging.info("Finally all tests done, going to compress test dir")
test_logs = os.path.join(str(self.path()), "./test_dir.tar")
self._compress_logs("{}/tests/integration".format(repo_path), test_logs)
logging.info("Compression finished")
result_path_dockerd_logs = os.path.join(str(self.path()), "dockerd.log")
if os.path.exists(result_path_dockerd_logs):
shutil.copy(DOCKERD_LOGS_PATH, result_path_dockerd_logs)
if counters["FAILED"] or counters["ERROR"]:
logging.info("Overall status failure, because we have tests in FAILED or ERROR state")
result_state = "failure"
else:
logging.info("Overall success!")
result_state = "success"
test_result = []
for state in ("ERROR", "FAILED", "PASSED", "SKIPPED", "FLAKY"):
if state == "PASSED":
text_state = "OK"
elif state == "FAILED":
text_state = "FAIL"
else:
text_state = state
test_result += [(c, text_state, "{:.2f}".format(tests_times[c])) for c in counters[state]]
failed_sum = len(counters['FAILED']) + len(counters['ERROR'])
status_text = "fail: {}, passed: {}, flaky: {}".format(failed_sum, len(counters['PASSED']), len(counters['FLAKY']))
if self.soft_deadline_time < time.time():
status_text = "Timeout, " + status_text
result_state = "failure"
counters['FLAKY'] = []
if not counters or sum(len(counter) for counter in counters.values()) == 0:
status_text = "No tests found for some reason! It's a bug"
result_state = "failure"
if '(memory)' in self.params['context_name']:
result_state = "success"
return result_state, status_text, test_result, [test_logs] + logs
def write_results(results_file, status_file, results, status):
with open(results_file, 'w') as f:
out = csv.writer(f, delimiter='\t')
out.writerows(results)
with open(status_file, 'w') as f:
out = csv.writer(f, delimiter='\t')
out.writerow(status)
if __name__ == "__main__":
logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s')
repo_path = os.environ.get("CLICKHOUSE_TESTS_REPO_PATH")
build_path = os.environ.get("CLICKHOUSE_TESTS_BUILD_PATH")
result_path = os.environ.get("CLICKHOUSE_TESTS_RESULT_PATH")
params_path = os.environ.get("CLICKHOUSE_TESTS_JSON_PARAMS_PATH")
params = json.loads(open(params_path, 'r').read())
runner = ClickhouseIntegrationTestsRunner(result_path, params)
logging.info("Running tests")
state, description, test_results, _ = runner.run_impl(repo_path, build_path)
logging.info("Tests finished")
status = (state, description)
out_results_file = os.path.join(str(runner.path()), "test_results.tsv")
out_status_file = os.path.join(str(runner.path()), "check_status.tsv")
write_results(out_results_file, out_status_file, test_results, status)
logging.info("Result written")