githubmirror/zulip

scripts/restart-server
337 строк13 KB
Alex Vandiver
restart-server: All realm Tornado ports need a restart when they change.
21 ноя 2025, 02:38
21 ноя 2025, 02:38f82cde7
#!/usr/bin/env python3
import contextlib
import json
import logging
import os
import pwd
import shlex
import subprocess
import sys
import time

sys.path.append(os.path.join(os.path.dirname(__file__), ".."))
from scripts.lib.setup_path import setup_path

setup_path()

from scripts.lib.supervisor import list_supervisor_processes
from scripts.lib.zulip_tools import (
    DEPLOYMENTS_DIR,
    ENDC,
    LOCK_DIR,
    OKGREEN,
    WARNING,
    get_config,
    get_config_file,
    get_tornado_ports,
    has_application_server,
    has_process_fts_updates,
    overwrite_symlink,
    start_arg_parser,
    su_to_zulip,
)

action = "restart"
if not sys.argv[0].endswith("restart-server"):
    action = "start"
verbing = action.title() + "ing"

logging.Formatter.converter = time.gmtime
logging.basicConfig(format=f"%(asctime)s {action}-server: %(message)s", level=logging.INFO)

parser = start_arg_parser(action=action, add_help=True)
args = parser.parse_args()

deploy_path = os.path.realpath(os.path.join(os.path.dirname(__file__), ".."))
os.chdir(deploy_path)

username = pwd.getpwuid(os.getuid()).pw_name
if username == "root":
    su_to_zulip()
elif username != "zulip":
    logging.error("Must be run as user 'zulip'.")
    sys.exit(1)

if os.environ.get("RUNNING_UNDER_CRON") and os.path.exists(LOCK_DIR):
    logging.info("Skipping cron-triggered restart during deploy.")
    sys.exit(1)

if args.tornado_reshard:
    tornado_reshard_files = ["/etc/zulip/sharding.json", "/etc/zulip/nginx_sharding_map.conf"]
    if not all(os.path.exists(f"{path}.tmp") for path in tornado_reshard_files):
        logging.info(
            "No resharding changes to apply!  Edit zulip.conf and run refresh-sharding-and-restart"
        )
        sys.exit(1)

if not args.skip_checks:
    logging.info("Running syntax and database checks")
    subprocess.check_call(["./manage.py", "check", "--database", "default"])

if args.fill_cache:
    logging.info("Filling memcached caches")
    subprocess.check_call(["./manage.py", "fill_memcached_caches", "--automated", "--skip-checks"])

current_symlink = os.path.join(DEPLOYMENTS_DIR, "current")
last_symlink = os.path.join(DEPLOYMENTS_DIR, "last")
change_symlink = os.readlink(current_symlink) != deploy_path
if change_symlink:
    overwrite_symlink(os.readlink(current_symlink), last_symlink)
    overwrite_symlink(deploy_path, current_symlink)

config_file = get_config_file()
tornado_ports = get_tornado_ports(config_file)
workers = []

if has_application_server():
    # Start by restarting the workers and similar processes, one at a
    # time.  Workers can always support processing events with old event
    # contents, but cannot necessarily understand events enqueued by a
    # newer Django process.  Restarting them one at a time, rather than
    # all-at-once, minimizes the downtime of each, and reduces startup
    # contention.
    #
    # For "start" or less-graceful circumstances, we don't need to
    # iterate; we'll stop all of them at once, and start them all later.
    # In those cases, using the glob form is faster -- but if we do need
    # to iterate, we need to expand the glob.
    if action == "start" or args.less_graceful:
        workers.append("zulip-workers:*")
    else:
        workers.extend(list_supervisor_processes(["zulip-workers:*"]))

    if has_application_server(once=True):
        # These used to be included in "zulip-workers:*"; since we may
        # be restarting an older version of Zulip, which has not
        # applied puppet to reload the new list of processes, only
        # stop them if they currently exist according to
        # `supervisorctl`.
        workers.extend(
            list_supervisor_processes(
                [
                    "zulip_deliver_scheduled_emails",
                    "zulip_deliver_scheduled_messages",
                ]
            )
        )

    # This is an optional service, so may or may not exist
    workers.extend(list_supervisor_processes(["zulip-katex"]))

    # This does have some Python code, which reads from settings.py,
    # so we need to restart it on every deploy.  A short outage during
    # the restart is fine, as clients will transparently retry.
    workers.append("zulip-tus")

    workers.extend(list_supervisor_processes(["zulip-email-server"]))

if has_process_fts_updates():
    workers.append("process-fts-updates")

# Before we start (re)starting main services, make sure to start any
# optional auxiliary services that we don't stop, but do expect to be
# running, and aren't currently.  These get new versions by getting
# updated supervisor files, and puppet restarts them -- so we never
# restart them in here, only start them.
aux_services = list_supervisor_processes(["go-camo", "smokescreen"], only_running=False)
if aux_services:
    subprocess.check_call(["supervisorctl", "start", *aux_services])

if args.only_django:
    workers = []
    check_services = ["zulip-django"]
else:
    check_services = [*workers, "zulip-django", "zulip-tornado:*"]

# If none of the workers nor the application servers are running, this
# is actually a "start," not a restart, which means we will defer
# workers to later.
running_services = list_supervisor_processes(check_services, only_running=True)
if action == "restart" and len(running_services) == 0:
    action = "start"
    verbing = "Starting"
elif action == "start":
    existing_services = list_supervisor_processes(check_services)
    if existing_services == running_services:
        # Check if the version we're trying to start is the version
        # that we would have started.  We do this via checking the CWD
        # of the oldest uwsgi worker.
        oldest_pid = subprocess.check_output(
            ["pgrep", "--oldest", "--full", "zulip-django uWSGI worker"], text=True
        ).strip()
        running_cwd = os.readlink(f"/proc/{oldest_pid}/cwd")
        if deploy_path != running_cwd:
            logging.warning("Mismatch between 'current' symlink and currently-running code!")
            logging.warning("Already-running deploy: %s", running_cwd)
            logging.warning("We would have started:  %s", deploy_path)
            logging.warning("To restart, call %s/scripts/restart-server", deploy_path)
        else:
            logging.info("Zulip is already started; nothing to do!")
        sys.exit(0)


def restart_or_start(service: str) -> None:
    our_verb = action
    existing_services = list_supervisor_processes([service])
    running_services = list_supervisor_processes([service], only_running=True)
    if our_verb == "restart" and len(running_services) == 0:
        our_verb = "start"
    elif our_verb == "start" and existing_services == running_services:
        logging.info("%s already started!", service)
        return
    subprocess.check_call(["supervisorctl", our_verb, service])


def ports_as_set(val: int | list[int]) -> frozenset[int]:
    return frozenset(val) if isinstance(val, list) else frozenset([val])


def update_tornado_sharding() -> list[int]:
    with open("/etc/zulip/sharding.json") as old_shard_fh:
        old_sharding = json.load(old_shard_fh)
    with open("/etc/zulip/sharding.json.tmp") as new_shard_fh:
        new_sharding = json.load(new_shard_fh)
    affected_tornados: set[int] = set()
    for realm in set().union(old_sharding["shard_map"], new_sharding["shard_map"]):
        old_ports = ports_as_set(old_sharding["shard_map"].get(realm, []))
        new_ports = ports_as_set(new_sharding["shard_map"].get(realm, []))
        if old_ports != new_ports:
            # A realm sharded across multiple ports gets requests at
            # random from nginx, and each does an X-Accel-Redirect to
            # the user's right Tornado instance. So all ports in the
            # set need a restart when we add or remove ports.
            affected_tornados |= old_ports | new_ports
    old_regex_set = {
        (regex, ports_as_set(ports)) for (regex, ports) in old_sharding["shard_regexes"]
    }
    new_regex_set = {
        (regex, ports_as_set(ports)) for (regex, ports) in new_sharding["shard_regexes"]
    }
    for regex, ports in old_regex_set ^ new_regex_set:
        affected_tornados |= ports

    tornado_reshard_files = ["/etc/zulip/sharding.json", "/etc/zulip/nginx_sharding_map.conf"]
    for path in tornado_reshard_files:
        os.rename(f"{path}.tmp", path)

    return list(affected_tornados)


if action == "restart" and len(workers) > 0:
    if args.less_graceful:
        # The less graceful form stops every worker now; we start them
        # back up at the end.
        logging.info("Stopping workers")
        subprocess.check_call(["supervisorctl", "stop", *workers])
    else:
        # We cannot pass all of these to one `supervisorctl restart`
        # because that takes them all down at once, waits until they are
        # all down, and then brings them back up; doing them sequentially
        # requires multiple `supervisorctl restart` calls.
        for worker in workers:
            logging.info("Restarting %s", worker)
            restart_or_start(worker)

if has_application_server():
    if args.tornado_reshard:
        affected_tornado_ports = update_tornado_sharding()
        logging.info("Tornado ports affected by this resharding: %s", affected_tornado_ports)
    else:
        affected_tornado_ports = tornado_ports
    if not args.only_django:
        # Next, we restart the Tornado processes sequentially, in order to
        # minimize downtime of the tornado service caused by too many
        # Python processes restarting at the same time, resulting in each
        # receiving insufficient priority.  This is important, because
        # Tornado being unavailable for too long is the main source of
        # user-visible downtime when we restart a Zulip server.  We do
        # this before restarting Django, in case there are new event types
        # which it will need to know how to deal with.
        if len(tornado_ports) > 1:
            for p in affected_tornado_ports:
                # Restart Tornado processes individually for a better rate of
                # restarts.  This also avoids behavior with restarting a whole
                # supervisord group where if any individual process is slow to
                # stop, the whole bundle stays stopped for an extended time.
                logging.info("%s Tornado process on port %s", verbing, p)
                restart_or_start(f"zulip-tornado:zulip-tornado-port-{p}")
        else:
            logging.info("%s Tornado process", verbing)
            restart_or_start("zulip-tornado:*")

    # Finally, restart the Django uWSGI processes.
    if (
        action == "restart"
        and not args.less_graceful
        and get_config(config_file, "application_server", "rolling_restart", False)
        and os.path.exists("/home/zulip/deployments/uwsgi-control")
    ):
        # See if it's currently running
        uwsgi_status = subprocess.run(
            ["supervisorctl", "status", "zulip-django"],
            stdout=subprocess.DEVNULL,
            check=False,
        )
        if uwsgi_status.returncode == 0:
            logging.info("Starting rolling restart of django server")
            with contextlib.suppress(FileNotFoundError):
                os.unlink("/var/lib/zulip/django-workers.ready")
            with open("/home/zulip/deployments/uwsgi-control", "w") as control_socket:
                # "c" is chain-reloading:
                # https://uwsgi-docs.readthedocs.io/en/latest/MasterFIFO.html#available-commands
                control_socket.write("c")
            n = 0
            while not os.path.exists("/var/lib/zulip/django-workers.ready"):
                time.sleep(1)
                n += 1
                if n % 5 == 0:
                    logging.info("...")
            logging.info("Chain reloading complete")
        else:
            logging.info("Starting django server")
            subprocess.check_call(["supervisorctl", "start", "zulip-django"])
    else:
        logging.info("%s django server", verbing)
        restart_or_start("zulip-django")

    using_sso = subprocess.check_output(["./scripts/get-django-setting", "USING_APACHE_SSO"])
    if using_sso.strip() == b"True":
        logging.info("Restarting Apache WSGI process...")
        subprocess.check_call(["pkill", "-x", "apache2", "-u", "zulip"])

# If we were doing this non-gracefully, or starting as opposed to
# restarting, we need to turn the workers (back) on.  There's no
# advantage to doing this not-all-at-once.
if (action == "start" or args.less_graceful) and not args.only_django:
    workers = list_supervisor_processes(workers, only_running=False)
    if workers:
        logging.info("Starting workers")
        subprocess.check_call(["supervisorctl", "start", *workers])

if (
    has_application_server()
    and not args.skip_client_reloads
    and not args.only_django
    and not args.tornado_reshard
):
    # All of the servers have been (re)started; now enqueue events in
    # the Tornado servers to tell clients to reload.
    logging.info("Sending reload events to clients in the background")
    subprocess.check_call(["./scripts/reload-clients", "--background"])

logging.info("Done!")
print(OKGREEN + f"Zulip {action}ed successfully!" + ENDC)

if change_symlink and "PWD" in os.environ:
    for symlink in [last_symlink, current_symlink]:
        if os.path.commonprefix([os.environ["PWD"], symlink]) == symlink:
            print(
                """
{}Your shell entered its current directory through a symlink:
  {}
which has now changed. Your shell will not see this change until you run:
  cd {}
to traverse the symlink again.{}
""".format(WARNING, symlink, shlex.quote(os.environ["PWD"]), ENDC),
                file=sys.stderr,
            )