Created using manage.py squashmigrations, with my patch to the Django
migration optimizer to correctly collapse
AddConstraints/RemoveConstraints operations.
This commit adds a new `group_size` field to the `DirectMessageGroup`
model, and backfills its value to each of the existing direct message
groups.
Fixes part of #25713
analytics_usercount_property_realm_id_end_time_591dbec1_idx, added
back in b7df84d5a8, makes this lookup actually indexed.
195defb031 rewrote this query in a way which stopped using the
index.
Instead of the PUSH_NOTIFICATIONS_BOUNCER_URL and
SUBMIT_USAGE_STATISTICS settings, we want servers to configure
individual ZULIP_SERVICE_* settings, while maintaining backward
compatibility with the old settings. Thus, if all the new
ZULIP_SERVICE_* are at their default False value, but the legacy
settings are activated, they need to be translated in computed_settings
to the modern way.
This commit renames the "Huddle" Django model class to
"DirectMessageGroup", while maintaining the same table --
"zerver_huddle".
Fixes part of #28640.
Migrate all `ids` of anything which does not have a foreign key from
the Message or UserMessage table (and would thus require walking
those) to be `bigint`. This is done by removing explicit
`BigAutoField`s, trading them for explicit `AutoField`s on the tables
to not be migrated, while updating `DEFAULT_AUTO_FIELD` to the new
default.
In general, the tables adjusted in this commit are small tables -- at
least compared to Messages and UserMessages.
Many-to-many tables without their own model class are adjusted by a
custom Operation, since they do not automatically pick up migrations
when `DEFAULT_AUTO_FIELD` changes[^1].
Note that this does multiple scans over tables to update foreign
keys[^2]. Large installs may wish to hand-optimize this using the
output of `./manage.py sqlmigrate` to join multiple `ALTER TABLE`
statements into one, to speed up the migration. This is unfortunately
not possible to do generically, as constraint names may differ between
installations.
This leaves the following primary keys as non-`bigint`:
- `auth_group.id`
- `auth_group_permissions.id`
- `auth_permission.id`
- `django_content_type.id`
- `django_migrations.id`
- `otp_static_staticdevice.id`
- `otp_static_statictoken.id`
- `otp_totp_totpdevice.id`
- `two_factor_phonedevice.id`
- `zerver_archivedmessage.id`
- `zerver_client.id`
- `zerver_message.id`
- `zerver_realm.id`
- `zerver_recipient.id`
- `zerver_userprofile.id`
[^1]: https://code.djangoproject.com/ticket/32674
[^2]: https://code.djangoproject.com/ticket/24203
This helps prevent wraparound on exceedingly large and old installs,
particularly Zulip Cloud. These are relatively simple migrations
since they are not referenced by any other tables; however, they are
quite large, and are actively used from Django by running servers,
making this not a migration which is possible to run without stopping
the server.
Use the escape hatch in the previous commit to temporarily pause
analytics writes while the migration happens. This should make the
migration transparent to users, at the small cost of an artificial dip
in statistics (specifically, to push notification counts, and unread
message counts) while the migration runs.
With `realm_active_humans` no longer dependent on the per-user rows,
there is no reason to preserve them -- any measure of "was a user
active" should look directly at the much richer RealmAuditLog. This
removes the bulk of the UserCount table, since the remaining rows all
require user interaction of some sort to produce rows.
This makes it no longer dependent on active_users_audit:is_bot:day,
which subsequent commits will make a RealmCount, not UserCount, query.
This folds the same behaviour of `active_users_audit` directly into
the query; however, only running over active users, using the index
from the earlier commit, and using the new `DISTINCT ON` formulation
make this a fast query compared to `active_users_audit:is_bot:day` +
the old `realm_active_humans::day`.
This produces a query which is more comprehensible, is 2x faster when
limited to a realm, and has equivalent speed when performing the full
table scan.
Updates the labels in the "Messages sent by client" analytics chart
for the user-agent/client names for the Flutter mobile app, which
can be "ZulipFlutter" or "ZulipMobile/flutter".
Fixes#28220.
This commit adds include_realm_default_subscriptions parameter
to the invite endpoints and the corresponding field in
PreregistrationUser and MultiuseInvite objects. This field will
be used to subscribe the new users to the default streams at the
time of account creation and not to the streams that were default
when sending the invite.
Fixes#29632.
The issue description explains this well:
We currently recalculate `currently_used_upload_space_bytes` every file
upload, by dint of calling `flush_used_upload_space_cache` on
save/delete, and then immediately calling
`user_profile.realm.currently_used_upload_space_bytes()` in
`notify_attachment_update`. Since this walks the Attachments table,
recalculating this can take seconds in large realms.
Switch this to using a CountStat, so we don't need to walk significant
chunks of the Attachment table when we upload an attachment. This will
also give us a historical daily graph of usage.
The previous implementation using Django's `get_or_create` for
`do_increment_logging_stat` involved two separate database queries,
potentially leading to race conditions.
Use an `ON CONFLICT ... DO UPDATE` (aka "upsert") query, which
eliminates race conditions and improves performance. This is mildly
complicated due to the different unique indexes across the various
tables, and the need for bug-for-bug compatibility with the previous
implementation.
Fixes#28947.
Co-authored-by: Alex Vandiver <alexmv@zulip.com>
The "invites" worker exists to do two things -- make a Confirmation
object, and send the outgoing email. Making the Confirmation object
in a background process from where the PreregistrationUser is created
temporarily leaves the PreregistrationUser in invalid state, and
results in 500's, and the user not immediately seeing the sent
invitation. That the "invites" worker also wants to create the
Confirmation object means that "resending" an invite invalidates the
URL in the previous email, which can be confusing to the user.
Moving the Confirmation creation to the same transaction solves both
of these issues, and leaves the "invites" worker with nothing to do
but send the email; as such, we remove it entirely, and use the
existing "email_senders" worker to send the invites. The volume of
invites is small enough that this will not affect other uses of that
worker.
Fixes: #21306Fixes: #24275
This prevents users from hammering the invitation endpoint, causing
races, and inviting more users than they should otherwise be allowed
to.
Doing this requires that we not raise InvitationError when we have
partially succeeded; that behaviour is left to the one callsite of
do_invite_users.
Reported by Lakshit Agarwal (@chiekosec).
Factor out the repeated pattern of taking a lock, or immediately
aborting with a message if it cannot be acquired. The exit code in
that situation is changed to be exit code 1, rather than the successful
0; we are likely missing new work since that process started.
We move the lockfiles to a common directory under `/srv/zulip-locks`
rather than muddy up `/home/zulip/deployments`.
Updates translated JsonableError strings that relate to streams
to use channel instead of stream. Separated from other error string
updates as this is a dense area of changes for this rename.
Part of stream to channel rename project.
Updates the labels for the "Messages sent by recipient type" chart
to use "channel", and updates the error message that would be sent
for the "messages_sent_by_stream" chart (that has not yet been
implemented) for a missing channel/stream.
Part of the stream to channel rename project.
Replaced HUDDLE attribute with DIRECT_MESSAGE_GROUP using VS Code search,
part of a general renaming of the object class.
Fixes part of #28640.
Co-authored-by: JohnLu2004 <JohnLu10212004@gmail.com>
Adds a re-usable lockfile_nonblocking helper to context_managers.
Relying on naive `os.mkdir` is not enough especially now that the
successful operation of this command is necessary for push notifications
to work for many servers.
We can't use `lockfile` context manager from
`zerver.lib.context_managers`, because we want the custom behavior of
failing if the lock can't be acquired, instead of waiting.
That's because if an instance of this gets stuck, we don't want to start
queueing up more processes waiting forever whenever the cronjob runs
again and fail->exit is preferrable instead.
LoggingCountStats with a daily duration and that are directly stored
on the RealmCount table (not via aggregation in process_count_stat),
can be in a state, after the hourly cron job to update analytics
counts, where the logged value will be live-updated later, because
the end time for the stat is still in the future.
As these logging counts are designed to be used on the self-hosted
installation for either debugging or rate limiting, sending these
partial/incomplete counts to the bouncer has low value.
The previous logic incorrectly used the server-level number of users
even when a (presumably smaller) realm-level count was available.
Fixes a bug introduced in 2e1ed4431a.
We don't have an index on `(server_id, id)`, and in any case, we have
a stronger guarantee that `remote_id` is time-sorted, from the
construction of the analytics tables, than that the `id`s given these
entries when uploaded are time-sorted.
Kenneth Rodrigues
Tim Abbott
roanster007
Anders Kaseorg
Alex Vandiver
Mateusz Mandera
Lauryn Menard
Sahil Batra
Mahhheshh
John Lu