upgrade: Add --skip-restart which preps but does not restart.

This adds a --skip-restart which makes `deployments/next` in a state
where it can be restarted into, but holds off on conducting that
restart.

This requires many of the same guarantees as `--skip-tornado`, in
terms of there being no Puppet or database schema changes between the
versions.  Enforce those with `--skip-restart`, and also broaden both
flags to prevent other, less common changes which nonetheless
potentially might affect the other deploy.
This commit is contained in:
Alex Vandiver 2022-05-02 18:45:15 -07:00 committed by Tim Abbott
parent 86a4e64726
commit 6337f17923
1 changed files with 89 additions and 62 deletions

View File

@ -81,6 +81,11 @@ restart_parser = start_arg_parser(action="restart", add_help=False)
parser = argparse.ArgumentParser(parents=[restart_parser])
parser.add_argument("deploy_path", metavar="deploy_path", help="Path to deployment directory")
parser.add_argument(
"--skip-restart",
action="store_true",
help="Configure, but do not restart into, the new version; aborts if any system-wide changes would happen.",
)
parser.add_argument("--skip-puppet", action="store_true", help="Skip doing puppet/apt upgrades.")
parser.add_argument("--skip-migrations", action="store_true", help="Skip doing migrations.")
parser.add_argument(
@ -104,18 +109,30 @@ parser.add_argument(
)
args = parser.parse_args()
if args.skip_tornado:
# There are two reasons we might want to do a "minimal change" which
# asserts puppet/database are unchanged: if we're setting up for a
# post-upgrade manual restart (--skip-restart), or if we're only
# restarting part of the system (--skip-tornado).
minimal_change = args.skip_restart or args.skip_tornado
if args.skip_restart and args.skip_tornado:
logging.warning("Ignored --skip-tornado; all upgrades with --skip-restart skip all restarts.")
args.skip_tornado = False
if minimal_change:
if args.skip_restart:
flagname = "--skip-restart"
else:
flagname = "--skip-tornado"
if args.less_graceful:
logging.warning("Ignored --less-graceful; --skip-tornado is always graceful.")
logging.warning("Ignored --less-graceful; %s is always graceful.", flagname)
args.less_graceful = False
if args.skip_migrations:
logging.warning(
"Ignored --skip-migrations; all upgrades with --skip-tornado assert no migrations."
"Ignored --skip-migrations; all upgrades with %s assert no migrations.", flagname
)
args.skip_migrations = False
if args.skip_puppet:
logging.warning(
"Ignored --skip-puppet; all upgrades with --skip-tornado assert no puppet changes."
"Ignored --skip-puppet; all upgrades with %s assert no puppet changes.", flagname
)
args.skip_puppet = False
@ -139,7 +156,7 @@ if os.path.exists("/var/lib/rabbitmq/.erlang.cookie"):
cookie_size = len(cookie_fh.readline())
else:
logging.info("No RabbitMQ erlang cookie found, not auditing RabbitMQ security.")
if args.skip_puppet and rabbitmq_dist_listen:
if (minimal_change or args.skip_puppet) and rabbitmq_dist_listen:
logging.error(
"RabbitMQ is publicly-accessible on %s; this is a security vulnerability!",
", ".join(rabbitmq_dist_listen),
@ -164,7 +181,7 @@ if args.skip_puppet and rabbitmq_dist_listen:
def shutdown_server() -> None:
global IS_SERVER_UP
if args.skip_tornado:
if minimal_change:
logging.info("Upgrade would require shutting down Zulip -- aborting!")
sys.exit(1)
@ -208,7 +225,7 @@ if glob.glob("/usr/share/postgresql/*/extension/tsearch_extras.control"):
)
subprocess.check_call(["apt-get", "remove", "-y", "postgresql-*-tsearch-extras"])
if not args.skip_puppet:
if not (minimal_change or args.skip_puppet):
logging.info("Upgrading system packages...")
subprocess.check_call(["apt-get", "update"])
subprocess.check_call(["apt-get", "-y", "upgrade"])
@ -261,6 +278,9 @@ class_renames = {
classes = re.split(r"\s*,\s*", get_config(config_file, "machine", "puppet_classes"))
new_classes = [class_renames.get(c, c) for c in classes if c != "zulip::base"]
if classes != new_classes:
if minimal_change:
logging.error("Would need to adjust puppet classes -- aborting!")
sys.exit(1)
logging.info("Adjusting Puppet classes for renames...")
subprocess.check_call(
[
@ -282,6 +302,9 @@ if os.path.exists(emoji_path):
emoji_data = f.read()
emoji_sha = hashlib.sha1(emoji_data).hexdigest()
if emoji_sha == "47033121dc20b376e0f86f4916969872ad22a293":
if minimal_change:
logging.error("Would need to delete images-google-64 -- aborting!")
sys.exit(1)
shutil.rmtree("/home/zulip/prod-static/generated/emoji/images-google-64")
# And then, building/installing the static assets.
@ -353,7 +376,7 @@ elif not args.skip_migrations:
if line_str.startswith("[ ]"):
migrations_needed = True
if args.skip_tornado and migrations_needed:
if minimal_change and migrations_needed:
logging.error("Would need to apply migrations -- aborting!")
sys.exit(1)
@ -380,16 +403,20 @@ if not args.skip_puppet and IS_SERVER_UP:
["./scripts/zulip-puppet-apply", "--noop", "--force"], stdout=subprocess.DEVNULL
)
if try_puppet.returncode == 0:
if args.skip_tornado:
if minimal_change:
logging.info("Verified no Puppet changes are necessary.")
else:
logging.info("No puppet changes found, skipping!")
args.skip_puppet = True
has_puppet_changes = False
elif args.skip_tornado:
elif minimal_change:
logging.error("Would need to apply puppet changes -- aborting!")
sys.exit(1)
if args.skip_restart:
logging.info("Successfully configured in %s!", deploy_path)
else:
# NOTE: Here begins the most likely critical period, where we may be
# shutting down the server; we should strive to minimize the number of
# steps that happen between here and the "Restarting Zulip" line