Persist partition|offset|user_id in the keyserver (#1226)

* Persist partition|offset|user_id in the keyserver

Required for a query API which will be used by the syncapi which
will be called when a `/sync` request comes in which will return
a list of user IDs of people who have changed their device keys
between two tokens.

* Add tests and fix maxOffset bug

* s/offset/log_offset/g because 'offset' is a reserved word in postgres

keyserver/keyserver.go

@@ -49,6 +49,7 @@ func NewInternalAPI(
 	keyChangeProducer := &producers.KeyChange{
 		Topic:    string(cfg.Kafka.Topics.OutputKeyChangeEvent),
 		Producer: producer,
+		DB:       db,
 	}
 	return &internal.KeyInternalAPI{
 		DB:         db,

keyserver/producers/keychange.go

@@ -15,10 +15,12 @@
 package producers
 
 import (
+	"context"
 	"encoding/json"
 
 	"github.com/Shopify/sarama"
 	"github.com/matrix-org/dendrite/keyserver/api"
+	"github.com/matrix-org/dendrite/keyserver/storage"
 	"github.com/sirupsen/logrus"
 )
 
@@ -26,6 +28,7 @@ import (
 type KeyChange struct {
 	Topic    string
 	Producer sarama.SyncProducer
+	DB       storage.Database
 }
 
 // ProduceKeyChanges creates new change events for each key
@@ -46,6 +49,10 @@ func (p *KeyChange) ProduceKeyChanges(keys []api.DeviceKeys) error {
 		if err != nil {
 			return err
 		}
+		err = p.DB.StoreKeyChange(context.Background(), partition, offset, key.UserID)
+		if err != nil {
+			return err
+		}
 		logrus.WithFields(logrus.Fields{
 			"user_id":   key.UserID,
 			"device_id": key.DeviceID,

keyserver/storage/interface.go

@@ -43,4 +43,12 @@ type Database interface {
 	// ClaimKeys based on the 3-uple of user_id, device_id and algorithm name. Returns the keys claimed. Returns no error if a key
 	// cannot be claimed or if none exist for this (user, device, algorithm), instead it is omitted from the returned slice.
 	ClaimKeys(ctx context.Context, userToDeviceToAlgorithm map[string]map[string]string) ([]api.OneTimeKeys, error)
+
+	// StoreKeyChange stores key change metadata after the change has been sent to Kafka. `userID` is the the user who has changed
+	// their keys in some way.
+	StoreKeyChange(ctx context.Context, partition int32, offset int64, userID string) error
+
+	// KeyChanges returns a list of user IDs who have modified their keys from the offset given.
+	// Returns the offset of the latest key change.
+	KeyChanges(ctx context.Context, partition int32, fromOffset int64) (userIDs []string, latestOffset int64, err error)
 }

keyserver/storage/postgres/key_changes_table.go

@@ -0,0 +1,97 @@
+// Copyright 2020 The Matrix.org Foundation C.I.C.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package postgres
+
+import (
+	"context"
+	"database/sql"
+
+	"github.com/matrix-org/dendrite/internal"
+	"github.com/matrix-org/dendrite/keyserver/storage/tables"
+)
+
+var keyChangesSchema = `
+-- Stores key change information about users. Used to determine when to send updated device lists to clients.
+CREATE TABLE IF NOT EXISTS keyserver_key_changes (
+	partition BIGINT NOT NULL,
+	log_offset BIGINT NOT NULL,
+    user_id TEXT NOT NULL,
+    CONSTRAINT keyserver_key_changes_unique UNIQUE (partition, log_offset)
+);
+`
+
+// Replace based on partition|offset - we should never insert duplicates unless the kafka logs are wiped.
+// Rather than falling over, just overwrite (though this will mean clients with an existing sync token will
+// miss out on updates). TODO: Ideally we would detect when kafka logs are purged then purge this table too.
+const upsertKeyChangeSQL = "" +
+	"INSERT INTO keyserver_key_changes (partition, log_offset, user_id)" +
+	" VALUES ($1, $2, $3)" +
+	" ON CONFLICT ON CONSTRAINT keyserver_key_changes_unique" +
+	" DO UPDATE SET user_id = $3"
+
+// select the highest offset for each user in the range. The grouping by user gives distinct entries and then we just
+// take the max offset value as the latest offset.
+const selectKeyChangesSQL = "" +
+	"SELECT user_id, MAX(log_offset) FROM keyserver_key_changes WHERE partition = $1 AND log_offset > $2 GROUP BY user_id"
+
+type keyChangesStatements struct {
+	db                   *sql.DB
+	upsertKeyChangeStmt  *sql.Stmt
+	selectKeyChangesStmt *sql.Stmt
+}
+
+func NewPostgresKeyChangesTable(db *sql.DB) (tables.KeyChanges, error) {
+	s := &keyChangesStatements{
+		db: db,
+	}
+	_, err := db.Exec(keyChangesSchema)
+	if err != nil {
+		return nil, err
+	}
+	if s.upsertKeyChangeStmt, err = db.Prepare(upsertKeyChangeSQL); err != nil {
+		return nil, err
+	}
+	if s.selectKeyChangesStmt, err = db.Prepare(selectKeyChangesSQL); err != nil {
+		return nil, err
+	}
+	return s, nil
+}
+
+func (s *keyChangesStatements) InsertKeyChange(ctx context.Context, partition int32, offset int64, userID string) error {
+	_, err := s.upsertKeyChangeStmt.ExecContext(ctx, partition, offset, userID)
+	return err
+}
+
+func (s *keyChangesStatements) SelectKeyChanges(
+	ctx context.Context, partition int32, fromOffset int64,
+) (userIDs []string, latestOffset int64, err error) {
+	rows, err := s.selectKeyChangesStmt.QueryContext(ctx, partition, fromOffset)
+	if err != nil {
+		return nil, 0, err
+	}
+	defer internal.CloseAndLogIfError(ctx, rows, "selectKeyChangesStmt: rows.close() failed")
+	for rows.Next() {
+		var userID string
+		var offset int64
+		if err := rows.Scan(&userID, &offset); err != nil {
+			return nil, 0, err
+		}
+		if offset > latestOffset {
+			latestOffset = offset
+		}
+		userIDs = append(userIDs, userID)
+	}
+	return
+}

keyserver/storage/postgres/storage.go

@@ -34,9 +34,14 @@ func NewDatabase(dbDataSourceName string, dbProperties sqlutil.DbProperties) (*s
 	if err != nil {
 		return nil, err
 	}
+	kc, err := NewPostgresKeyChangesTable(db)
+	if err != nil {
+		return nil, err
+	}
 	return &shared.Database{
 		DB:               db,
 		OneTimeKeysTable: otk,
 		DeviceKeysTable:  dk,
+		KeyChangesTable:  kc,
 	}, nil
 }

keyserver/storage/shared/storage.go

@@ -28,6 +28,7 @@ type Database struct {
 	DB               *sql.DB
 	OneTimeKeysTable tables.OneTimeKeys
 	DeviceKeysTable  tables.DeviceKeys
+	KeyChangesTable  tables.KeyChanges
 }
 
 func (d *Database) ExistingOneTimeKeys(ctx context.Context, userID, deviceID string, keyIDsWithAlgorithms []string) (map[string]json.RawMessage, error) {
@@ -72,3 +73,11 @@ func (d *Database) ClaimKeys(ctx context.Context, userToDeviceToAlgorithm map[st
 	})
 	return result, err
 }
+
+func (d *Database) StoreKeyChange(ctx context.Context, partition int32, offset int64, userID string) error {
+	return d.KeyChangesTable.InsertKeyChange(ctx, partition, offset, userID)
+}
+
+func (d *Database) KeyChanges(ctx context.Context, partition int32, fromOffset int64) (userIDs []string, latestOffset int64, err error) {
+	return d.KeyChangesTable.SelectKeyChanges(ctx, partition, fromOffset)
+}

keyserver/storage/sqlite3/key_changes_table.go

@@ -0,0 +1,98 @@
+// Copyright 2020 The Matrix.org Foundation C.I.C.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package sqlite3
+
+import (
+	"context"
+	"database/sql"
+
+	"github.com/matrix-org/dendrite/internal"
+	"github.com/matrix-org/dendrite/keyserver/storage/tables"
+)
+
+var keyChangesSchema = `
+-- Stores key change information about users. Used to determine when to send updated device lists to clients.
+CREATE TABLE IF NOT EXISTS keyserver_key_changes (
+	partition BIGINT NOT NULL,
+	offset BIGINT NOT NULL,
+	-- The key owner
+	user_id TEXT NOT NULL,
+	UNIQUE (partition, offset)
+);
+`
+
+// Replace based on partition|offset - we should never insert duplicates unless the kafka logs are wiped.
+// Rather than falling over, just overwrite (though this will mean clients with an existing sync token will
+// miss out on updates). TODO: Ideally we would detect when kafka logs are purged then purge this table too.
+const upsertKeyChangeSQL = "" +
+	"INSERT INTO keyserver_key_changes (partition, offset, user_id)" +
+	" VALUES ($1, $2, $3)" +
+	" ON CONFLICT (partition, offset)" +
+	" DO UPDATE SET user_id = $3"
+
+// select the highest offset for each user in the range. The grouping by user gives distinct entries and then we just
+// take the max offset value as the latest offset.
+const selectKeyChangesSQL = "" +
+	"SELECT user_id, MAX(offset) FROM keyserver_key_changes WHERE partition = $1 AND offset > $2 GROUP BY user_id"
+
+type keyChangesStatements struct {
+	db                   *sql.DB
+	upsertKeyChangeStmt  *sql.Stmt
+	selectKeyChangesStmt *sql.Stmt
+}
+
+func NewSqliteKeyChangesTable(db *sql.DB) (tables.KeyChanges, error) {
+	s := &keyChangesStatements{
+		db: db,
+	}
+	_, err := db.Exec(keyChangesSchema)
+	if err != nil {
+		return nil, err
+	}
+	if s.upsertKeyChangeStmt, err = db.Prepare(upsertKeyChangeSQL); err != nil {
+		return nil, err
+	}
+	if s.selectKeyChangesStmt, err = db.Prepare(selectKeyChangesSQL); err != nil {
+		return nil, err
+	}
+	return s, nil
+}
+
+func (s *keyChangesStatements) InsertKeyChange(ctx context.Context, partition int32, offset int64, userID string) error {
+	_, err := s.upsertKeyChangeStmt.ExecContext(ctx, partition, offset, userID)
+	return err
+}
+
+func (s *keyChangesStatements) SelectKeyChanges(
+	ctx context.Context, partition int32, fromOffset int64,
+) (userIDs []string, latestOffset int64, err error) {
+	rows, err := s.selectKeyChangesStmt.QueryContext(ctx, partition, fromOffset)
+	if err != nil {
+		return nil, 0, err
+	}
+	defer internal.CloseAndLogIfError(ctx, rows, "selectKeyChangesStmt: rows.close() failed")
+	for rows.Next() {
+		var userID string
+		var offset int64
+		if err := rows.Scan(&userID, &offset); err != nil {
+			return nil, 0, err
+		}
+		if offset > latestOffset {
+			latestOffset = offset
+		}
+		userIDs = append(userIDs, userID)
+	}
+	return
+}

keyserver/storage/sqlite3/storage.go

@@ -37,9 +37,14 @@ func NewDatabase(dataSourceName string) (*shared.Database, error) {
 	if err != nil {
 		return nil, err
 	}
+	kc, err := NewSqliteKeyChangesTable(db)
+	if err != nil {
+		return nil, err
+	}
 	return &shared.Database{
 		DB:               db,
 		OneTimeKeysTable: otk,
 		DeviceKeysTable:  dk,
+		KeyChangesTable:  kc,
 	}, nil
 }

keyserver/storage/storage_test.go

@@ -0,0 +1,57 @@
+package storage
+
+import (
+	"context"
+	"reflect"
+	"testing"
+)
+
+var ctx = context.Background()
+
+func MustNotError(t *testing.T, err error) {
+	t.Helper()
+	if err == nil {
+		return
+	}
+	t.Fatalf("operation failed: %s", err)
+}
+
+func TestKeyChanges(t *testing.T) {
+	db, err := NewDatabase("file::memory:", nil)
+	if err != nil {
+		t.Fatalf("Failed to NewDatabase: %s", err)
+	}
+	MustNotError(t, db.StoreKeyChange(ctx, 0, 0, "@alice:localhost"))
+	MustNotError(t, db.StoreKeyChange(ctx, 0, 1, "@bob:localhost"))
+	MustNotError(t, db.StoreKeyChange(ctx, 0, 2, "@charlie:localhost"))
+	userIDs, latest, err := db.KeyChanges(ctx, 0, 1)
+	if err != nil {
+		t.Fatalf("Failed to KeyChanges: %s", err)
+	}
+	if latest != 2 {
+		t.Fatalf("KeyChanges: got latest=%d want 2", latest)
+	}
+	if !reflect.DeepEqual(userIDs, []string{"@charlie:localhost"}) {
+		t.Fatalf("KeyChanges: wrong user_ids: %v", userIDs)
+	}
+}
+
+func TestKeyChangesNoDupes(t *testing.T) {
+	db, err := NewDatabase("file::memory:", nil)
+	if err != nil {
+		t.Fatalf("Failed to NewDatabase: %s", err)
+	}
+	MustNotError(t, db.StoreKeyChange(ctx, 0, 0, "@alice:localhost"))
+	MustNotError(t, db.StoreKeyChange(ctx, 0, 1, "@alice:localhost"))
+	MustNotError(t, db.StoreKeyChange(ctx, 0, 2, "@alice:localhost"))
+	userIDs, latest, err := db.KeyChanges(ctx, 0, 0)
+	if err != nil {
+		t.Fatalf("Failed to KeyChanges: %s", err)
+	}
+	if latest != 2 {
+		t.Fatalf("KeyChanges: got latest=%d want 2", latest)
+	}
+	if !reflect.DeepEqual(userIDs, []string{"@alice:localhost"}) {
+		t.Fatalf("KeyChanges: wrong user_ids: %v", userIDs)
+	}
+}

keyserver/storage/tables/interface.go

@@ -35,3 +35,8 @@ type DeviceKeys interface {
 	InsertDeviceKeys(ctx context.Context, keys []api.DeviceKeys) error
 	SelectBatchDeviceKeys(ctx context.Context, userID string, deviceIDs []string) ([]api.DeviceKeys, error)
 }
+
+type KeyChanges interface {
+	InsertKeyChange(ctx context.Context, partition int32, offset int64, userID string) error
+	SelectKeyChanges(ctx context.Context, partition int32, fromOffset int64) (userIDs []string, latestOffset int64, err error)
+}