Fetching contributors…
Cannot retrieve contributors at this time
128 lines (103 sloc) 5.37 KB
import os
import shutil
import subprocess
import tempfile
from argparse import ArgumentParser, RawTextHelpFormatter
from typing import Any
from import CommandError
from zerver.lib.export import do_export_realm, \
from import ZulipBaseCommand
class Command(ZulipBaseCommand):
help = """Exports all data from a Zulip realm
This command exports all significant data from a Zulip realm. The
result can be imported using the `./ import` command.
Things that are exported:
* All user-accessible data in the Zulip database (Messages,
Streams, UserMessages, RealmEmoji, etc.)
* Copies of all uploaded files and avatar images along with
metadata needed to restore them even in the ab
Things that are not exported:
* Confirmation and PreregistrationUser (transient tables)
* Sessions (everyone will need to login again post-export)
* Users' passwords and API keys (users will need to use SSO or reset password)
* Mobile tokens for APNS/GCM (users will need to reconnect their mobile devices)
* ScheduledEmail (Not relevant on a new server)
* RemoteZulipServer (Unlikely to be migrated)
* third_party_api_results cache (this means rerending all old
messages could be expensive)
Things that will break as a result of the export:
* Passwords will not be transferred. They will all need to go
through the password reset flow to obtain a new password (unless
they intend to only use e.g. Google Auth).
* Users will need to logout and re-login to the Zulip desktop and
mobile apps. The apps now all have an option on the login page
where you can specify which Zulip server to use; your users
should enter <domain name>.
* All bots will stop working since they will be pointing to the
wrong server URL, and all users' API keys have been rotated as
part of the migration. So to re-enable your integrations, you
will need to direct your integrations at the new server.
Usually this means updating the URL and the bots' API keys. You
can see a list of all the bots that have been configured for
your realm on the `/#organization` page, and use that list to
make sure you migrate them all.
The proper procedure for using this to export a realm is as follows:
* Use `./ deactivate_realm` to deactivate the realm, so
nothing happens in the realm being exported during the export
* Use `./ export` to export the realm, producing a data
* Transfer the tarball to the new server and unpack it.
* Use `./ import` to import the realm
* Use `./ reactivate_realm` to reactivate the realm, so
users can login again.
* Inform the users about the things broken above.
We recommend testing by exporting without having deactivated the
realm first, to make sure you have the procedure right and
minimize downtime.
Performance: In one test, the tool exported a realm with hundreds
of users and ~1M messages of history with --threads=1 in about 3
hours of serial runtime (goes down to ~50m with --threads=6 on a
machine with 8 CPUs). Importing that same data set took about 30
minutes. But this will vary a lot depending on the average number
of recipients of messages in the realm, hardware, etc."""
# Fix support for multi-line usage
def create_parser(self, *args: Any, **kwargs: Any) -> ArgumentParser:
parser = super().create_parser(*args, **kwargs)
parser.formatter_class = RawTextHelpFormatter
return parser
def add_arguments(self, parser: ArgumentParser) -> None:
help='Directory to write exported data to.')
help='Threads to use in exporting UserMessage objects in parallel')
self.add_realm_args(parser, True)
def handle(self, *args: Any, **options: Any) -> None:
realm = self.get_realm(options)
assert realm is not None # Should be ensured by parser
output_dir = options["output_dir"]
if output_dir is None:
output_dir = tempfile.mkdtemp(prefix="/tmp/zulip-export-")
output_dir = os.path.realpath(os.path.expanduser(output_dir))
if os.path.exists(output_dir):
print("Exporting realm %s" % (realm.string_id,))
num_threads = int(options['threads'])
if num_threads < 1:
raise CommandError('You must have at least one thread.')
do_export_realm(realm, output_dir, threads=num_threads)
print("Finished exporting to %s; tarring" % (output_dir,))
tarball_path = output_dir.rstrip('/') + '.tar.gz'
subprocess.check_call(["tar", "-czf", tarball_path, os.path.basename(output_dir)])
print("Tarball written to %s" % (tarball_path,))