diff options
author | Richard van der Hoff <1389908+richvdh@users.noreply.github.com> | 2019-08-22 10:42:59 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2019-08-22 10:42:59 +0100 |
commit | 119aa31b105705390e87f87186f787b32e04ba21 (patch) | |
tree | 6b254acd81c9c11942d556aaa3ec5efd6bdbcdc0 /synapse/storage | |
parent | Improve error msg when key-fetch fails (#5896) (diff) | |
download | synapse-119aa31b105705390e87f87186f787b32e04ba21.tar.xz |
Servlet to purge old rooms (#5845)
Diffstat (limited to 'synapse/storage')
-rw-r--r-- | synapse/storage/events.py | 137 |
1 files changed, 137 insertions, 0 deletions
diff --git a/synapse/storage/events.py b/synapse/storage/events.py index 6fcfa4d789..5a95c36a8b 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -2181,6 +2181,143 @@ class EventsStore( return to_delete, to_dedelta + def purge_room(self, room_id): + """Deletes all record of a room + + Args: + room_id (str): + """ + + return self.runInteraction("purge_room", self._purge_room_txn, room_id) + + def _purge_room_txn(self, txn, room_id): + # first we have to delete the state groups states + logger.info("[purge] removing %s from state_groups_state", room_id) + + txn.execute( + """ + DELETE FROM state_groups_state WHERE state_group IN ( + SELECT state_group FROM events JOIN event_to_state_groups USING(event_id) + WHERE events.room_id=? + ) + """, + (room_id,), + ) + + # ... and the state group edges + logger.info("[purge] removing %s from state_group_edges", room_id) + + txn.execute( + """ + DELETE FROM state_group_edges WHERE state_group IN ( + SELECT state_group FROM events JOIN event_to_state_groups USING(event_id) + WHERE events.room_id=? + ) + """, + (room_id,), + ) + + # ... and the state groups + logger.info("[purge] removing %s from state_groups", room_id) + + txn.execute( + """ + DELETE FROM state_groups WHERE id IN ( + SELECT state_group FROM events JOIN event_to_state_groups USING(event_id) + WHERE events.room_id=? + ) + """, + (room_id,), + ) + + # and then tables which lack an index on room_id but have one on event_id + for table in ( + "event_auth", + "event_edges", + "event_push_actions_staging", + "event_reference_hashes", + "event_relations", + "event_to_state_groups", + "redactions", + "rejections", + "state_events", + ): + logger.info("[purge] removing %s from %s", room_id, table) + + txn.execute( + """ + DELETE FROM %s WHERE event_id IN ( + SELECT event_id FROM events WHERE room_id=? + ) + """ + % (table,), + (room_id,), + ) + + # and finally, the tables with an index on room_id (or no useful index) + for table in ( + "current_state_events", + "event_backward_extremities", + "event_forward_extremities", + "event_json", + "event_push_actions", + "event_search", + "events", + "group_rooms", + "public_room_list_stream", + "receipts_graph", + "receipts_linearized", + "room_aliases", + "room_depth", + "room_memberships", + "room_state", + "room_stats", + "room_stats_earliest_token", + "rooms", + "stream_ordering_to_exterm", + "topics", + "users_in_public_rooms", + "users_who_share_private_rooms", + # no useful index, but let's clear them anyway + "appservice_room_list", + "e2e_room_keys", + "event_push_summary", + "pusher_throttle", + "group_summary_rooms", + "local_invites", + "room_account_data", + "room_tags", + ): + logger.info("[purge] removing %s from %s", room_id, table) + txn.execute("DELETE FROM %s WHERE room_id=?" % (table,), (room_id,)) + + # Other tables we do NOT need to clear out: + # + # - blocked_rooms + # This is important, to make sure that we don't accidentally rejoin a blocked + # room after it was purged + # + # - user_directory + # This has a room_id column, but it is unused + # + + # Other tables that we might want to consider clearing out include: + # + # - event_reports + # Given that these are intended for abuse management my initial + # inclination is to leave them in place. + # + # - current_state_delta_stream + # - ex_outlier_stream + # - room_tags_revisions + # The problem with these is that they are largeish and there is no room_id + # index on them. In any case we should be clearing out 'stream' tables + # periodically anyway (#5888) + + # TODO: we could probably usefully do a bunch of cache invalidation here + + logger.info("[purge] done") + @defer.inlineCallbacks def is_event_after(self, event_id1, event_id2): """Returns True if event_id1 is after event_id2 in the stream |