ClickHouse/tests/integration/runner

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import subprocess
import os
import getpass
import glob
import argparse
import logging
import signal
import subprocess
import sys
import string
import random


def random_str(length=6):
    alphabet = string.ascii_lowercase + string.digits
    return "".join(random.SystemRandom().choice(alphabet) for _ in range(length))


CUR_FILE_DIR = os.path.dirname(os.path.realpath(__file__))
DEFAULT_CLICKHOUSE_ROOT = os.path.abspath(os.path.join(CUR_FILE_DIR, "../../"))
CURRENT_WORK_DIR = os.getcwd()
VOLUME_NAME = "clickhouse_integration_tests"
CONTAINER_NAME = f"{VOLUME_NAME}_{random_str()}"

CONFIG_DIR_IN_REPO = "programs/server"
INTEGRATION_DIR_IN_REPO = "tests/integration"
SRC_DIR_IN_REPO = "src"

DIND_INTEGRATION_TESTS_IMAGE_NAME = "clickhouse/integration-tests-runner"


def check_args_and_update_paths(args):
    if args.clickhouse_root:
        if not os.path.isabs(args.clickhouse_root):
            CLICKHOUSE_ROOT = os.path.abspath(args.clickhouse_root)
        else:
            CLICKHOUSE_ROOT = args.clickhouse_root
    else:
        logging.info(
            "ClickHouse root is not set. Will use %s" % (DEFAULT_CLICKHOUSE_ROOT)
        )
        CLICKHOUSE_ROOT = DEFAULT_CLICKHOUSE_ROOT

    if not os.path.isabs(args.binary):
        args.binary = os.path.abspath(os.path.join(CURRENT_WORK_DIR, args.binary))

    if not args.odbc_bridge_binary:
        args.odbc_bridge_binary = os.path.join(
            os.path.dirname(args.binary), "clickhouse-odbc-bridge"
        )
    elif not os.path.isabs(args.odbc_bridge_binary):
        args.odbc_bridge_binary = os.path.abspath(
            os.path.join(CURRENT_WORK_DIR, args.odbc_bridge_binary)
        )

    if not args.library_bridge_binary:
        args.library_bridge_binary = os.path.join(
            os.path.dirname(args.binary), "clickhouse-library-bridge"
        )
    elif not os.path.isabs(args.library_bridge_binary):
        args.library_bridge_binary = os.path.abspath(
            os.path.join(CURRENT_WORK_DIR, args.library_bridge_binary)
        )

    if args.base_configs_dir:
        if not os.path.isabs(args.base_configs_dir):
            args.base_configs_dir = os.path.abspath(
                os.path.join(CURRENT_WORK_DIR, args.base_configs_dir)
            )
    else:
        args.base_configs_dir = os.path.abspath(
            os.path.join(CLICKHOUSE_ROOT, CONFIG_DIR_IN_REPO)
        )
        logging.info(
            "Base configs dir is not set. Will use %s" % (args.base_configs_dir)
        )

    if args.cases_dir:
        if not os.path.isabs(args.cases_dir):
            args.cases_dir = os.path.abspath(
                os.path.join(CURRENT_WORK_DIR, args.cases_dir)
            )
    else:
        args.cases_dir = os.path.abspath(
            os.path.join(CLICKHOUSE_ROOT, INTEGRATION_DIR_IN_REPO)
        )
        logging.info("Cases dir is not set. Will use %s" % (args.cases_dir))

    if args.src_dir:
        if not os.path.isabs(args.src_dir):
            args.src_dir = os.path.abspath(os.path.join(CURRENT_WORK_DIR, args.src_dir))
    else:
        args.src_dir = os.path.abspath(os.path.join(CLICKHOUSE_ROOT, SRC_DIR_IN_REPO))
        logging.info("src dir is not set. Will use %s" % (args.src_dir))

    logging.info(
        "base_configs_dir: {}, binary: {}, cases_dir: {} ".format(
            args.base_configs_dir, args.binary, args.cases_dir
        )
    )

    for path in [
        args.binary,
        args.odbc_bridge_binary,
        args.library_bridge_binary,
        args.base_configs_dir,
        args.cases_dir,
        CLICKHOUSE_ROOT,
    ]:
        if not os.path.exists(path):
            raise Exception("Path {} doesn't exist".format(path))

    if args.dockerd_volume:
        if not os.path.isabs(args.dockerd_volume):
            args.src_dir = os.path.abspath(
                os.path.join(CURRENT_WORK_DIR, args.dockerd_volume)
            )

    if (not os.path.exists(os.path.join(args.base_configs_dir, "config.xml"))) and (
        not os.path.exists(os.path.join(args.base_configs_dir, "config.yaml"))
    ):
        raise Exception(
            "No config.xml or config.yaml in {}".format(args.base_configs_dir)
        )

    if (not os.path.exists(os.path.join(args.base_configs_dir, "users.xml"))) and (
        not os.path.exists(os.path.join(args.base_configs_dir, "users.yaml"))
    ):
        raise Exception(
            "No users.xml or users.yaml in {}".format(args.base_configs_dir)
        )


def docker_kill_handler_handler(signum, frame):
    subprocess.check_call(
        'docker ps --all --quiet --filter name={name} --format="{{{{.ID}}}}"'.format(
            name=CONTAINER_NAME
        ),
        shell=True,
    )
    raise KeyboardInterrupt("Killed by Ctrl+C")


signal.signal(signal.SIGINT, docker_kill_handler_handler)

# Integration tests runner should allow to run tests on several versions of ClickHouse.
# Integration tests should be portable.
# To run integration tests following artfacts should be sufficient:
#   - clickhouse binaries (env CLICKHOUSE_TESTS_SERVER_BIN_PATH or --binary arg)
#   - clickhouse default configs(config.xml, users.xml) from same version as binary (env CLICKHOUSE_TESTS_BASE_CONFIG_DIR or --base-configs-dir arg)
#   - odbc bridge binary (env CLICKHOUSE_TESTS_ODBC_BRIDGE_BIN_PATH or --odbc-bridge-binary arg)
#   - library bridge binary (env CLICKHOUSE_TESTS_LIBRARY_BRIDGE_BIN_PATH or --library-bridge-binary)
#   - tests/integration directory with all test cases and configs (env CLICKHOUSE_TESTS_INTEGRATION_PATH or --cases-dir)
#
# 1) --clickhouse-root is only used to determine other paths on default places
# 2) path of runner script is used to determine paths for trivial case, when we run it from repository

if __name__ == "__main__":
    logging.basicConfig(
        level=logging.INFO,
        format="%(asctime)s [ %(process)d ] %(levelname)s : %(message)s (%(filename)s:%(lineno)s, %(funcName)s)",
    )
    parser = argparse.ArgumentParser(description="ClickHouse integration tests runner")

    parser.add_argument(
        "--binary",
        default=os.environ.get(
            "CLICKHOUSE_TESTS_SERVER_BIN_PATH",
            os.environ.get("CLICKHOUSE_TESTS_CLIENT_BIN_PATH", "/usr/bin/clickhouse"),
        ),
        help="Path to clickhouse binary. For example /usr/bin/clickhouse",
    )

    parser.add_argument(
        "--odbc-bridge-binary",
        default=os.environ.get("CLICKHOUSE_TESTS_ODBC_BRIDGE_BIN_PATH", ""),
        help="Path to clickhouse-odbc-bridge binary. Defaults to clickhouse-odbc-bridge in the same dir as clickhouse.",
    )

    parser.add_argument(
        "--library-bridge-binary",
        default=os.environ.get("CLICKHOUSE_TESTS_LIBRARY_BRIDGE_BIN_PATH", ""),
        help="Path to clickhouse-library-bridge binary. Defaults to clickhouse-library-bridge in the same dir as clickhouse.",
    )

    parser.add_argument(
        "--base-configs-dir",
        default=os.environ.get("CLICKHOUSE_TESTS_BASE_CONFIG_DIR"),
        help="Path to clickhouse base configs directory with config.xml/users.xml",
    )

    parser.add_argument(
        "--cases-dir",
        default=os.environ.get("CLICKHOUSE_TESTS_INTEGRATION_PATH"),
        help="Path to integration tests cases and configs directory. For example tests/integration in repository",
    )

    parser.add_argument(
        "--src-dir",
        default=os.environ.get("CLICKHOUSE_SRC_DIR"),
        help="Path to the 'src' directory in repository. Used to provide schemas (e.g. *.proto) for some tests when those schemas are located in the 'src' directory",
    )

    parser.add_argument(
        "--clickhouse-root",
        help="Path to repository root folder. Used to take configuration from repository default paths.",
    )

    parser.add_argument(
        "--command",
        default="",
        help="Set it to run some other command in container (for example bash)",
    )

    parser.add_argument(
        "--disable-net-host",
        action="store_true",
        default=False,
        help="Don't use net host in parent docker container",
    )

    parser.add_argument(
        "--network",
        help="Set network driver for runnner container (defaults to `host`)",
    )

    parser.add_argument(
        "--docker-image-version",
        default="latest",
        help="Version of docker image which runner will use to run tests",
    )

    parser.add_argument(
        "--docker-compose-images-tags",
        action="append",
        help="Set non-default tags for images used in docker compose recipes(yandex/my_container:my_tag)",
    )

    parser.add_argument(
        "-n", "--parallel", action="store", dest="parallel", help="Parallelism"
    )

    parser.add_argument(
        "--no-random", action="store", dest="no_random", help="Disable tests order randomization"
    )

    parser.add_argument(
        "--pre-pull", action="store_true", default=False, dest="pre_pull", help="Pull images for docker_compose before all other actions"
    )

    parser.add_argument(
        "-t",
        "--tests_list",
        action="store",
        nargs="+",
        default=[],
        dest="tests_list",
        help="List of tests to run",
    )

    parser.add_argument(
        "-k",
        "--keyword_expression",
        action="store",
        dest="keyword_expression",
        help="pytest keyword expression",
    )

    parser.add_argument(
        "--tmpfs",
        action="store_true",
        default=False,
        dest="tmpfs",
        help="Use tmpfs for dockerd files",
    )

    parser.add_argument(
        "--cleanup-containers",
        action="store_true",
        default=False,
        dest="cleanup_containers",
        help="Remove all running containers on test session start",
    )

    parser.add_argument(
        "--dockerd-volume-dir",
        action="store",
        dest="dockerd_volume",
        help="Bind volume to this dir to use for dockerd files",
    )

    parser.add_argument("pytest_args", nargs="*", help="args for pytest command")

    args = parser.parse_args()

    check_args_and_update_paths(args)

    parallel_args = ""
    if args.parallel:
        parallel_args += "--dist=loadfile"
        parallel_args += " -n {}".format(args.parallel)

    rand_args = ""
    # if not args.no_random:
    #     rand_args += f"--random-seed={os.getpid()}"


    net = ""
    if args.network:
        net = "--net={}".format(args.network)
    elif not args.disable_net_host:
        net = "--net=host"

    env_tags = ""

    if args.docker_compose_images_tags is not None:
        for img_tag in args.docker_compose_images_tags:
            [image, tag] = img_tag.split(":")
            if image == "clickhouse/mysql-golang-client":
                env_tags += "-e {}={} ".format("DOCKER_MYSQL_GOLANG_CLIENT_TAG", tag)
            elif image == "clickhouse/dotnet-client":
                env_tags += "-e {}={} ".format("DOCKER_DOTNET_CLIENT_TAG", tag)
            elif image == "clickhouse/mysql-java-client":
                env_tags += "-e {}={} ".format("DOCKER_MYSQL_JAVA_CLIENT_TAG", tag)
            elif image == "clickhouse/mysql-js-client":
                env_tags += "-e {}={} ".format("DOCKER_MYSQL_JS_CLIENT_TAG", tag)
            elif image == "clickhouse/mysql-php-client":
                env_tags += "-e {}={} ".format("DOCKER_MYSQL_PHP_CLIENT_TAG", tag)
            elif image == "clickhouse/postgresql-java-client":
                env_tags += "-e {}={} ".format("DOCKER_POSTGRESQL_JAVA_CLIENT_TAG", tag)
            elif image == "clickhouse/integration-test":
                env_tags += "-e {}={} ".format("DOCKER_BASE_TAG", tag)
            elif image == "clickhouse/kerberized-hadoop":
                env_tags += "-e {}={} ".format("DOCKER_KERBERIZED_HADOOP_TAG", tag)
            elif image == "clickhouse/kerberos-kdc":
                env_tags += "-e {}={} ".format("DOCKER_KERBEROS_KDC_TAG", tag)
            else:
                logging.info("Unknown image %s" % (image))

    # create named volume which will be used inside to store images and other docker related files,
    # to avoid redownloading it every time
    #
    # should be removed manually when not needed
    dockerd_internal_volume = ""
    if args.tmpfs:
        dockerd_internal_volume = "--tmpfs /var/lib/docker -e DOCKER_RAMDISK=true"
    elif args.dockerd_volume:
        dockerd_internal_volume = (
            "--mount type=bind,source={},target=/var/lib/docker".format(
                args.dockerd_volume
            )
        )
    else:
        try:
            subprocess.check_call(
                f"docker volume create {VOLUME_NAME}_volume", shell=True
            )
        except Exception as ex:
            print("Volume creation failed, probably it already exists, exception", ex)
        # TODO: this part cleans out stale volumes produced by container name
        # randomizer, we should remove it after Sep 2022
        try:
            subprocess.check_call(
                f"docker volume ls -q | grep '{VOLUME_NAME}_.*_volume' | xargs --no-run-if-empty docker volume rm",
                shell=True,
            )
        except Exception as ex:
            print("Probably, some stale volumes still there, just continue:", ex)
        # TODO END
        dockerd_internal_volume = f"--volume={VOLUME_NAME}_volume:/var/lib/docker"

    # If enabled we kill and remove containers before pytest session run.
    env_cleanup = ""
    if args.cleanup_containers:
        env_cleanup = "-e PYTEST_CLEANUP_CONTAINERS=1"
    # enable tty mode & interactive for docker if we have real tty
    tty = ""
    if sys.stdout.isatty() and sys.stdin.isatty():
        tty = "-it"

    # Remove old logs.
    for old_log_path in glob.glob(args.cases_dir + "/pytest*.log"):
        os.remove(old_log_path)

    if args.keyword_expression:
        args.pytest_args += ["-k", args.keyword_expression]

    cmd_base = "docker run {net} {tty} --rm --name {name} --privileged \
        --volume={odbc_bridge_bin}:/clickhouse-odbc-bridge --volume={bin}:/clickhouse \
        --volume={library_bridge_bin}:/clickhouse-library-bridge \
        --volume={base_cfg}:/clickhouse-config --volume={cases_dir}:/ClickHouse/tests/integration \
        --volume={src_dir}/Server/grpc_protos:/ClickHouse/src/Server/grpc_protos \
        --volume=/run:/run/host:ro \
        {dockerd_internal_volume} -e DOCKER_CLIENT_TIMEOUT=300 -e COMPOSE_HTTP_TIMEOUT=600 \
        -e XTABLES_LOCKFILE=/run/host/xtables.lock \
        -e PYTHONUNBUFFERED=1 \
        {env_tags} {env_cleanup} -e PYTEST_OPTS='{parallel} {opts} {tests_list} {rand} -vvv' {img}".format(
        net=net,
        tty=tty,
        bin=args.binary,
        odbc_bridge_bin=args.odbc_bridge_binary,
        library_bridge_bin=args.library_bridge_binary,
        base_cfg=args.base_configs_dir,
        cases_dir=args.cases_dir,
        src_dir=args.src_dir,
        env_tags=env_tags,
        env_cleanup=env_cleanup,
        parallel=parallel_args,
        rand=rand_args,
        opts=" ".join(args.pytest_args).replace("'", "\\'"),
        tests_list=" ".join(args.tests_list),
        dockerd_internal_volume=dockerd_internal_volume,
        img=DIND_INTEGRATION_TESTS_IMAGE_NAME + ":" + args.docker_image_version,
        name=CONTAINER_NAME,
    )

    cmd = cmd_base + " " +  args.command
    cmd_pre_pull = cmd_base + " find /compose -name docker_compose_*.yml -exec docker-compose -f '{}' pull \;"

    containers = subprocess.check_output(
        f"docker ps --all --quiet --filter name={CONTAINER_NAME} --format={{{{.ID}}}}",
        shell=True,
        universal_newlines=True,
    ).splitlines()
    if containers:
        print(f"Trying to kill containers name={CONTAINER_NAME} ids={containers}")
        subprocess.check_call(f"docker kill {' '.join(containers)}", shell=True)
        print(f"Containers {containers} killed")

    if args.pre_pull:
        print(("Running pre pull as: '" + cmd_pre_pull + "'."))
        subprocess.check_call(cmd_pre_pull, shell=True)

    print(("Running pytest container as: '" + cmd + "'."))
    subprocess.check_call(cmd, shell=True)