0
0
Fork 0
mirror of https://github.com/matrix-org/dendrite synced 2024-11-18 15:50:52 +01:00
dendrite/roomserver/internal/perform/perform_backfill.go

610 lines
22 KiB
Go
Raw Normal View History

// 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 perform
import (
"context"
"fmt"
"github.com/getsentry/sentry-go"
federationAPI "github.com/matrix-org/dendrite/federationapi/api"
"github.com/matrix-org/dendrite/internal/eventutil"
"github.com/matrix-org/dendrite/roomserver/api"
"github.com/matrix-org/dendrite/roomserver/auth"
"github.com/matrix-org/dendrite/roomserver/internal/helpers"
"github.com/matrix-org/dendrite/roomserver/storage"
"github.com/matrix-org/dendrite/roomserver/types"
"github.com/matrix-org/gomatrixserverlib"
"github.com/matrix-org/util"
"github.com/sirupsen/logrus"
)
// the max number of servers to backfill from per request. If this is too low we may fail to backfill when
// we could've from another server. If this is too high we may take far too long to successfully backfill
// as we try dead servers.
const maxBackfillServers = 5
type Backfiller struct {
ServerName gomatrixserverlib.ServerName
DB storage.Database
FSAPI federationAPI.RoomserverFederationAPI
KeyRing gomatrixserverlib.JSONVerifier
// The servers which should be preferred above other servers when backfilling
PreferServers []gomatrixserverlib.ServerName
}
// PerformBackfill implements api.RoomServerQueryAPI
func (r *Backfiller) PerformBackfill(
ctx context.Context,
request *api.PerformBackfillRequest,
response *api.PerformBackfillResponse,
) error {
// if we are requesting the backfill then we need to do a federation hit
// TODO: we could be more sensible and fetch as many events we already have then request the rest
// which is what the syncapi does already.
if request.ServerName == r.ServerName {
return r.backfillViaFederation(ctx, request, response)
}
// someone else is requesting the backfill, try to service their request.
var err error
var front []string
// The limit defines the maximum number of events to retrieve, so it also
// defines the highest number of elements in the map below.
visited := make(map[string]bool, request.Limit)
// this will include these events which is what we want
front = request.PrevEventIDs()
info, err := r.DB.RoomInfo(ctx, request.RoomID)
if err != nil {
return err
}
if info == nil || info.IsStub {
return fmt.Errorf("PerformBackfill: missing room info for room %s", request.RoomID)
}
// Scan the event tree for events to send back.
Roomserver/federation input refactor (#2104) * Put federation client functions into their own file * Look for missing auth events in RS input * Remove retrieveMissingAuthEvents from federation API * Logging * Sorta transplanted the code over * Use event origin failing all else * Don't get stuck on mutexes: * Add verifier * Don't mark state events with zero snapshot NID as not existing * Check missing state if not an outlier before storing the event * Reject instead of soft-fail, don't copy roominfo so much * Use synchronous contexts, limit time to fetch missing events * Clean up some commented out bits * Simplify `/send` endpoint significantly * Submit async * Report errors on sending to RS input * Set max payload in NATS to 16MB * Tweak metrics * Add `workerForRoom` for tidiness * Try skipping unmarshalling errors for RespMissingEvents * Track missing prev events separately to avoid calculating state when not possible * Tweak logic around checking missing state * Care about state when checking missing prev events * Don't check missing state for create events * Try that again * Handle create events better * Send create room events as new * Use given event kind when sending auth/state events * Revert "Use given event kind when sending auth/state events" This reverts commit 089d64d271b5fca8c104e1554711187420dbebca. * Only search for missing prev events or state for new events * Tweaks * We only have missing prev if we don't supply state * Room version tweaks * Allow async inputs again * Apply backpressure to consumers/synchronous requests to hopefully stop things being overwhelmed * Set timeouts on roomserver input tasks (need to decide what timeout makes sense) * Use work queue policy, deliver all on restart * Reduce chance of duplicates being sent by NATS * Limit the number of servers we attempt to reduce backpressure * Some review comment fixes * Tidy up a couple things * Don't limit servers, randomise order using map * Some context refactoring * Update gmsl * Don't resend create events * Set stateIDs length correctly or else the roomserver thinks there are missing events when there aren't * Exclude our own servername * Try backing off servers * Make excluding self behaviour optional * Exclude self from g_m_e * Update sytest-whitelist * Update consumers for the roomserver output stream * Remember to send outliers for state returned from /gme * Make full HTTP tests less upsetti * Remove 'If a device list update goes missing, the server resyncs on the next one' from the sytest blacklist * Remove debugging test * Fix blacklist again, remove unnecessary duplicate context * Clearer contexts, don't use background in case there's something happening there * Don't queue up events more than once in memory * Correctly identify create events when checking for state * Fill in gaps again in /gme code * Remove `AuthEventIDs` from `InputRoomEvent` * Remove stray field Co-authored-by: Kegan Dougal <kegan@matrix.org>
2022-01-27 15:29:14 +01:00
resultNIDs, err := helpers.ScanEventTree(ctx, r.DB, info, front, visited, request.Limit, request.ServerName)
if err != nil {
return err
}
// Retrieve events from the list that was filled previously. If we fail to get
// events from the database then attempt once to get them from federation instead.
var loadedEvents []*gomatrixserverlib.Event
loadedEvents, err = helpers.LoadEvents(ctx, r.DB, resultNIDs)
if err != nil {
if _, ok := err.(types.MissingEventError); ok {
return r.backfillViaFederation(ctx, request, response)
}
return err
}
for _, event := range loadedEvents {
response.Events = append(response.Events, event.Headered(info.RoomVersion))
}
return err
}
func (r *Backfiller) backfillViaFederation(ctx context.Context, req *api.PerformBackfillRequest, res *api.PerformBackfillResponse) error {
info, err := r.DB.RoomInfo(ctx, req.RoomID)
if err != nil {
return err
}
if info == nil || info.IsStub {
return fmt.Errorf("backfillViaFederation: missing room info for room %s", req.RoomID)
}
requester := newBackfillRequester(r.DB, r.FSAPI, r.ServerName, req.BackwardsExtremities, r.PreferServers)
// Request 100 items regardless of what the query asks for.
// We don't want to go much higher than this.
// We can't honour exactly the limit as some sytests rely on requesting more for tests to pass
// (so we don't need to hit /state_ids which the test has no listener for)
// Specifically the test "Outbound federation can backfill events"
events, err := gomatrixserverlib.RequestBackfill(
ctx, requester,
r.KeyRing, req.RoomID, info.RoomVersion, req.PrevEventIDs(), 100)
if err != nil {
return err
}
logrus.WithField("room_id", req.RoomID).Infof("backfilled %d events", len(events))
// persist these new events - auth checks have already been done
roomNID, backfilledEventMap := persistEvents(ctx, r.DB, events)
if err != nil {
return err
}
for _, ev := range backfilledEventMap {
// now add state for these events
stateIDs, ok := requester.eventIDToBeforeStateIDs[ev.EventID()]
if !ok {
// this should be impossible as all events returned must have pass Step 5 of the PDU checks
// which requires a list of state IDs.
logrus.WithError(err).WithField("event_id", ev.EventID()).Error("backfillViaFederation: failed to find state IDs for event which passed auth checks")
continue
}
var entries []types.StateEntry
if entries, err = r.DB.StateEntriesForEventIDs(ctx, stateIDs); err != nil {
// attempt to fetch the missing events
r.fetchAndStoreMissingEvents(ctx, info.RoomVersion, requester, stateIDs)
// try again
entries, err = r.DB.StateEntriesForEventIDs(ctx, stateIDs)
if err != nil {
logrus.WithError(err).WithField("event_id", ev.EventID()).Error("backfillViaFederation: failed to get state entries for event")
return err
}
}
var beforeStateSnapshotNID types.StateSnapshotNID
if beforeStateSnapshotNID, err = r.DB.AddState(ctx, roomNID, nil, entries); err != nil {
logrus.WithError(err).WithField("event_id", ev.EventID()).Error("backfillViaFederation: failed to persist state entries to get snapshot nid")
return err
}
if err = r.DB.SetState(ctx, ev.EventNID, beforeStateSnapshotNID); err != nil {
logrus.WithError(err).WithField("event_id", ev.EventID()).Error("backfillViaFederation: failed to persist snapshot nid")
}
}
// TODO: update backwards extremities, as that should be moved from syncapi to roomserver at some point.
res.Events = events
return nil
}
// fetchAndStoreMissingEvents does a best-effort fetch and store of missing events specified in stateIDs. Returns no error as it is just
// best effort.
func (r *Backfiller) fetchAndStoreMissingEvents(ctx context.Context, roomVer gomatrixserverlib.RoomVersion,
backfillRequester *backfillRequester, stateIDs []string) {
servers := backfillRequester.servers
// work out which are missing
nidMap, err := r.DB.EventNIDs(ctx, stateIDs)
if err != nil {
util.GetLogger(ctx).WithError(err).Warn("cannot query missing events")
return
}
missingMap := make(map[string]*gomatrixserverlib.HeaderedEvent) // id -> event
for _, id := range stateIDs {
if _, ok := nidMap[id]; !ok {
missingMap[id] = nil
}
}
util.GetLogger(ctx).Infof("Fetching %d missing state events (from %d possible servers)", len(missingMap), len(servers))
// fetch the events from federation. Loop the servers first so if we find one that works we stick with them
for _, srv := range servers {
for id, ev := range missingMap {
if ev != nil {
continue // already found
}
logger := util.GetLogger(ctx).WithField("server", srv).WithField("event_id", id)
res, err := r.FSAPI.GetEvent(ctx, srv, id)
if err != nil {
logger.WithError(err).Warn("failed to get event from server")
continue
}
loader := gomatrixserverlib.NewEventsLoader(roomVer, r.KeyRing, backfillRequester, backfillRequester.ProvideEvents, false)
result, err := loader.LoadAndVerify(ctx, res.PDUs, gomatrixserverlib.TopologicalOrderByPrevEvents)
if err != nil {
logger.WithError(err).Warn("failed to load and verify event")
continue
}
logger.Infof("returned %d PDUs which made events %+v", len(res.PDUs), result)
for _, res := range result {
switch err := res.Error.(type) {
case nil:
case gomatrixserverlib.SignatureErr:
// The signature of the event might not be valid anymore, for example if
// the key ID was reused with a different signature.
logger.WithError(err).Errorf("event failed PDU checks, storing anyway")
case gomatrixserverlib.AuthChainErr, gomatrixserverlib.AuthRulesErr:
logger.WithError(err).Warn("event failed PDU checks")
continue
default:
logger.WithError(err).Warn("event failed PDU checks")
continue
}
missingMap[id] = res.Event
}
}
}
var newEvents []*gomatrixserverlib.HeaderedEvent
for _, ev := range missingMap {
if ev != nil {
newEvents = append(newEvents, ev)
}
}
util.GetLogger(ctx).Infof("Persisting %d new events", len(newEvents))
persistEvents(ctx, r.DB, newEvents)
}
// backfillRequester implements gomatrixserverlib.BackfillRequester
type backfillRequester struct {
db storage.Database
fsAPI federationAPI.RoomserverFederationAPI
thisServer gomatrixserverlib.ServerName
preferServer map[gomatrixserverlib.ServerName]bool
bwExtrems map[string][]string
// per-request state
servers []gomatrixserverlib.ServerName
eventIDToBeforeStateIDs map[string][]string
eventIDMap map[string]*gomatrixserverlib.Event
}
func newBackfillRequester(
db storage.Database, fsAPI federationAPI.RoomserverFederationAPI, thisServer gomatrixserverlib.ServerName,
bwExtrems map[string][]string, preferServers []gomatrixserverlib.ServerName,
) *backfillRequester {
preferServer := make(map[gomatrixserverlib.ServerName]bool)
for _, p := range preferServers {
preferServer[p] = true
}
return &backfillRequester{
db: db,
fsAPI: fsAPI,
thisServer: thisServer,
eventIDToBeforeStateIDs: make(map[string][]string),
eventIDMap: make(map[string]*gomatrixserverlib.Event),
bwExtrems: bwExtrems,
preferServer: preferServer,
}
}
func (b *backfillRequester) StateIDsBeforeEvent(ctx context.Context, targetEvent *gomatrixserverlib.HeaderedEvent) ([]string, error) {
b.eventIDMap[targetEvent.EventID()] = targetEvent.Unwrap()
if ids, ok := b.eventIDToBeforeStateIDs[targetEvent.EventID()]; ok {
return ids, nil
}
if len(targetEvent.PrevEventIDs()) == 0 && targetEvent.Type() == "m.room.create" && targetEvent.StateKeyEquals("") {
util.GetLogger(ctx).WithField("room_id", targetEvent.RoomID()).Info("Backfilled to the beginning of the room")
b.eventIDToBeforeStateIDs[targetEvent.EventID()] = []string{}
return nil, nil
}
// if we have exactly 1 prev event and we know the state of the room at that prev event, then just roll forward the prev event.
// Else, we have to hit /state_ids because either we don't know the state at all at this event (new backwards extremity) or
// we don't know the result of state res to merge forks (2 or more prev_events)
if len(targetEvent.PrevEventIDs()) == 1 {
prevEventID := targetEvent.PrevEventIDs()[0]
prevEvent, ok := b.eventIDMap[prevEventID]
if !ok {
goto FederationHit
}
prevEventStateIDs, ok := b.eventIDToBeforeStateIDs[prevEventID]
if !ok {
goto FederationHit
}
newStateIDs := b.calculateNewStateIDs(targetEvent.Unwrap(), prevEvent, prevEventStateIDs)
if newStateIDs != nil {
b.eventIDToBeforeStateIDs[targetEvent.EventID()] = newStateIDs
return newStateIDs, nil
}
// else we failed to calculate the new state, so fallthrough
}
FederationHit:
var lastErr error
logrus.WithField("event_id", targetEvent.EventID()).Info("Requesting /state_ids at event")
for _, srv := range b.servers { // hit any valid server
c := gomatrixserverlib.FederatedStateProvider{
FedClient: b.fsAPI,
RememberAuthEvents: false,
Server: srv,
}
res, err := c.StateIDsBeforeEvent(ctx, targetEvent)
if err != nil {
lastErr = err
continue
}
b.eventIDToBeforeStateIDs[targetEvent.EventID()] = res
return res, nil
}
sentry.CaptureException(lastErr) // temporary to see if we might need to raise the server limit
return nil, lastErr
}
func (b *backfillRequester) calculateNewStateIDs(targetEvent, prevEvent *gomatrixserverlib.Event, prevEventStateIDs []string) []string {
newStateIDs := prevEventStateIDs[:]
if prevEvent.StateKey() == nil {
// state is the same as the previous event
b.eventIDToBeforeStateIDs[targetEvent.EventID()] = newStateIDs
return newStateIDs
}
missingState := false // true if we are missing the info for a state event ID
foundEvent := false // true if we found a (type, state_key) match
// find which state ID to replace, if any
for i, id := range newStateIDs {
ev, ok := b.eventIDMap[id]
if !ok {
missingState = true
continue
}
// The state IDs BEFORE the target event are the state IDs BEFORE the prev_event PLUS the prev_event itself
if ev.Type() == prevEvent.Type() && ev.StateKeyEquals(*prevEvent.StateKey()) {
newStateIDs[i] = prevEvent.EventID()
foundEvent = true
break
}
}
if !foundEvent && !missingState {
// we can be certain that this is new state
newStateIDs = append(newStateIDs, prevEvent.EventID())
foundEvent = true
}
if foundEvent {
b.eventIDToBeforeStateIDs[targetEvent.EventID()] = newStateIDs
return newStateIDs
}
return nil
}
func (b *backfillRequester) StateBeforeEvent(ctx context.Context, roomVer gomatrixserverlib.RoomVersion,
event *gomatrixserverlib.HeaderedEvent, eventIDs []string) (map[string]*gomatrixserverlib.Event, error) {
// try to fetch the events from the database first
events, err := b.ProvideEvents(roomVer, eventIDs)
if err != nil {
// non-fatal, fallthrough
logrus.WithError(err).Info("Failed to fetch events")
} else {
logrus.Infof("Fetched %d/%d events from the database", len(events), len(eventIDs))
if len(events) == len(eventIDs) {
result := make(map[string]*gomatrixserverlib.Event)
for i := range events {
result[events[i].EventID()] = events[i]
b.eventIDMap[events[i].EventID()] = events[i]
}
return result, nil
}
}
var lastErr error
for _, srv := range b.servers {
c := gomatrixserverlib.FederatedStateProvider{
FedClient: b.fsAPI,
RememberAuthEvents: false,
Server: srv,
}
result, err := c.StateBeforeEvent(ctx, roomVer, event, eventIDs)
if err != nil {
lastErr = err
continue
}
for eventID, ev := range result {
b.eventIDMap[eventID] = ev
}
return result, nil
}
sentry.CaptureException(lastErr) // temporary to see if we might need to raise the server limit
return nil, lastErr
}
// ServersAtEvent is called when trying to determine which server to request from.
// It returns a list of servers which can be queried for backfill requests. These servers
// will be servers that are in the room already. The entries at the beginning are preferred servers
// and will be tried first. An empty list will fail the request.
func (b *backfillRequester) ServersAtEvent(ctx context.Context, roomID, eventID string) []gomatrixserverlib.ServerName {
// eventID will be a prev_event ID of a backwards extremity, meaning we will not have a database entry for it. Instead, use
// its successor, so look it up.
successor := ""
FindSuccessor:
for sucID, prevEventIDs := range b.bwExtrems {
for _, pe := range prevEventIDs {
if pe == eventID {
successor = sucID
break FindSuccessor
}
}
}
if successor == "" {
logrus.WithField("event_id", eventID).Error("ServersAtEvent: failed to find successor of this event to determine room state")
return nil
}
eventID = successor
// getMembershipsBeforeEventNID requires a NID, so retrieving the NID for
// the event is necessary.
NIDs, err := b.db.EventNIDs(ctx, []string{eventID})
if err != nil {
logrus.WithField("event_id", eventID).WithError(err).Error("ServersAtEvent: failed to get event NID for event")
return nil
}
info, err := b.db.RoomInfo(ctx, roomID)
if err != nil {
logrus.WithError(err).WithField("room_id", roomID).Error("ServersAtEvent: failed to get RoomInfo for room")
return nil
}
if info == nil || info.IsStub {
logrus.WithField("room_id", roomID).Error("ServersAtEvent: failed to get RoomInfo for room, room is missing")
return nil
}
Roomserver/federation input refactor (#2104) * Put federation client functions into their own file * Look for missing auth events in RS input * Remove retrieveMissingAuthEvents from federation API * Logging * Sorta transplanted the code over * Use event origin failing all else * Don't get stuck on mutexes: * Add verifier * Don't mark state events with zero snapshot NID as not existing * Check missing state if not an outlier before storing the event * Reject instead of soft-fail, don't copy roominfo so much * Use synchronous contexts, limit time to fetch missing events * Clean up some commented out bits * Simplify `/send` endpoint significantly * Submit async * Report errors on sending to RS input * Set max payload in NATS to 16MB * Tweak metrics * Add `workerForRoom` for tidiness * Try skipping unmarshalling errors for RespMissingEvents * Track missing prev events separately to avoid calculating state when not possible * Tweak logic around checking missing state * Care about state when checking missing prev events * Don't check missing state for create events * Try that again * Handle create events better * Send create room events as new * Use given event kind when sending auth/state events * Revert "Use given event kind when sending auth/state events" This reverts commit 089d64d271b5fca8c104e1554711187420dbebca. * Only search for missing prev events or state for new events * Tweaks * We only have missing prev if we don't supply state * Room version tweaks * Allow async inputs again * Apply backpressure to consumers/synchronous requests to hopefully stop things being overwhelmed * Set timeouts on roomserver input tasks (need to decide what timeout makes sense) * Use work queue policy, deliver all on restart * Reduce chance of duplicates being sent by NATS * Limit the number of servers we attempt to reduce backpressure * Some review comment fixes * Tidy up a couple things * Don't limit servers, randomise order using map * Some context refactoring * Update gmsl * Don't resend create events * Set stateIDs length correctly or else the roomserver thinks there are missing events when there aren't * Exclude our own servername * Try backing off servers * Make excluding self behaviour optional * Exclude self from g_m_e * Update sytest-whitelist * Update consumers for the roomserver output stream * Remember to send outliers for state returned from /gme * Make full HTTP tests less upsetti * Remove 'If a device list update goes missing, the server resyncs on the next one' from the sytest blacklist * Remove debugging test * Fix blacklist again, remove unnecessary duplicate context * Clearer contexts, don't use background in case there's something happening there * Don't queue up events more than once in memory * Correctly identify create events when checking for state * Fill in gaps again in /gme code * Remove `AuthEventIDs` from `InputRoomEvent` * Remove stray field Co-authored-by: Kegan Dougal <kegan@matrix.org>
2022-01-27 15:29:14 +01:00
stateEntries, err := helpers.StateBeforeEvent(ctx, b.db, info, NIDs[eventID])
if err != nil {
logrus.WithField("event_id", eventID).WithError(err).Error("ServersAtEvent: failed to load state before event")
return nil
}
// possibly return all joined servers depending on history visiblity
memberEventsFromVis, err := joinEventsFromHistoryVisibility(ctx, b.db, roomID, stateEntries, b.thisServer)
if err != nil {
logrus.WithError(err).Error("ServersAtEvent: failed calculate servers from history visibility rules")
return nil
}
logrus.Infof("ServersAtEvent including %d current events from history visibility", len(memberEventsFromVis))
// Retrieve all "m.room.member" state events of "join" membership, which
// contains the list of users in the room before the event, therefore all
// the servers in it at that moment.
memberEvents, err := helpers.GetMembershipsAtState(ctx, b.db, stateEntries, true)
if err != nil {
logrus.WithField("event_id", eventID).WithError(err).Error("ServersAtEvent: failed to get memberships before event")
return nil
}
memberEvents = append(memberEvents, memberEventsFromVis...)
// Store the server names in a temporary map to avoid duplicates.
serverSet := make(map[gomatrixserverlib.ServerName]bool)
for _, event := range memberEvents {
serverSet[event.Origin()] = true
}
var servers []gomatrixserverlib.ServerName
for server := range serverSet {
if server == b.thisServer {
continue
}
if b.preferServer[server] { // insert at the front
servers = append([]gomatrixserverlib.ServerName{server}, servers...)
} else { // insert at the back
servers = append(servers, server)
}
}
if len(servers) > maxBackfillServers {
servers = servers[:maxBackfillServers]
}
b.servers = servers
return servers
}
// Backfill performs a backfill request to the given server.
// https://matrix.org/docs/spec/server_server/latest#get-matrix-federation-v1-backfill-roomid
func (b *backfillRequester) Backfill(ctx context.Context, server gomatrixserverlib.ServerName, roomID string,
limit int, fromEventIDs []string) (gomatrixserverlib.Transaction, error) {
tx, err := b.fsAPI.Backfill(ctx, server, roomID, limit, fromEventIDs)
return tx, err
}
func (b *backfillRequester) ProvideEvents(roomVer gomatrixserverlib.RoomVersion, eventIDs []string) ([]*gomatrixserverlib.Event, error) {
ctx := context.Background()
nidMap, err := b.db.EventNIDs(ctx, eventIDs)
if err != nil {
logrus.WithError(err).WithField("event_ids", eventIDs).Error("Failed to find events")
return nil, err
}
eventNIDs := make([]types.EventNID, len(nidMap))
i := 0
for _, nid := range nidMap {
eventNIDs[i] = nid
i++
}
eventsWithNids, err := b.db.Events(ctx, eventNIDs)
if err != nil {
logrus.WithError(err).WithField("event_nids", eventNIDs).Error("Failed to load events")
return nil, err
}
events := make([]*gomatrixserverlib.Event, len(eventsWithNids))
for i := range eventsWithNids {
events[i] = eventsWithNids[i].Event
}
return events, nil
}
// joinEventsFromHistoryVisibility returns all CURRENTLY joined members if our server can read the room history
// TODO: Long term we probably want a history_visibility table which stores eventNID | visibility_enum so we can just
// pull all events and then filter by that table.
func joinEventsFromHistoryVisibility(
ctx context.Context, db storage.Database, roomID string, stateEntries []types.StateEntry,
thisServer gomatrixserverlib.ServerName) ([]types.Event, error) {
var eventNIDs []types.EventNID
for _, entry := range stateEntries {
// Filter the events to retrieve to only keep the membership events
if entry.EventTypeNID == types.MRoomHistoryVisibilityNID && entry.EventStateKeyNID == types.EmptyStateKeyNID {
eventNIDs = append(eventNIDs, entry.EventNID)
break
}
}
// Get all of the events in this state
stateEvents, err := db.Events(ctx, eventNIDs)
if err != nil {
return nil, err
}
events := make([]*gomatrixserverlib.Event, len(stateEvents))
for i := range stateEvents {
events[i] = stateEvents[i].Event
}
// Can we see events in the room?
canSeeEvents := auth.IsServerAllowed(thisServer, true, events)
if !canSeeEvents {
logrus.Infof("ServersAtEvent history not visible to us: %s", auth.HistoryVisibilityForRoom(events))
return nil, nil
}
// get joined members
info, err := db.RoomInfo(ctx, roomID)
if err != nil {
return nil, err
}
joinEventNIDs, err := db.GetMembershipEventNIDsForRoom(ctx, info.RoomNID, true, false)
if err != nil {
return nil, err
}
return db.Events(ctx, joinEventNIDs)
}
func persistEvents(ctx context.Context, db storage.Database, events []*gomatrixserverlib.HeaderedEvent) (types.RoomNID, map[string]types.Event) {
var roomNID types.RoomNID
var eventNID types.EventNID
backfilledEventMap := make(map[string]types.Event)
for j, ev := range events {
nidMap, err := db.EventNIDs(ctx, ev.AuthEventIDs())
if err != nil { // this shouldn't happen as RequestBackfill already found them
logrus.WithError(err).WithField("auth_events", ev.AuthEventIDs()).Error("Failed to find one or more auth events")
continue
}
authNids := make([]types.EventNID, len(nidMap))
i := 0
for _, nid := range nidMap {
authNids[i] = nid
i++
}
var redactedEventID string
var redactionEvent *gomatrixserverlib.Event
eventNID, roomNID, _, redactionEvent, redactedEventID, err = db.StoreEvent(ctx, ev.Unwrap(), authNids, false)
if err != nil {
logrus.WithError(err).WithField("event_id", ev.EventID()).Error("Failed to persist event")
continue
}
// If storing this event results in it being redacted, then do so.
// It's also possible for this event to be a redaction which results in another event being
// redacted, which we don't care about since we aren't returning it in this backfill.
if redactedEventID == ev.EventID() {
eventToRedact := ev.Unwrap()
Ristretto cache (#2563) * Try Ristretto cache * Tweak * It's beautiful * Update GMSL * More strict keyable interface * Fix that some more * Make less panicky * Don't enforce mutability checks for now * Determine mutability using deep equality * Tweaks * Namespace keys * Make federation caches mutable * Update cost estimation, add metric * Update GMSL * Estimate cost for metrics better * Reduce counters a bit * Try caching events * Some guards * Try again * Try this * Use separate caches for hopefully better hash distribution * Fix bug with admitting events into cache * Try to fix bugs * Check nil * Try that again * Preserve order jeezo this is messy * thanks VS Code for doing exactly the wrong thing * Try this again * Be more specific * aaaaargh * One more time * That might be better * Stronger sorting * Cache expiries, async publishing of EDUs * Put it back * Use a shared cache again * Cost estimation fixes * Update ristretto * Reduce counters a bit * Clean up a bit * Update GMSL * 1GB * Configurable cache sizees * Tweaks * Add `config.DataUnit` for specifying friendly cache sizes * Various tweaks * Update GMSL * Add back some lazy loading caching * Include key in cost * Include key in cost * Tweak max age handling, config key name * Only register prometheus metrics if requested * Review comments @S7evinK * Don't return errors when creating caches (it is better just to crash since otherwise we'll `nil`-pointer exception everywhere) * Review comments * Update sample configs * Update GHA Workflow * Update Complement images to Go 1.18 * Remove the cache test from the federation API as we no longer guarantee immediate cache admission * Don't check the caches in the renewal test * Possibly fix the upgrade tests * Update to matrix-org/gomatrixserverlib#322 * Update documentation to refer to Go 1.18
2022-07-11 15:31:31 +02:00
if err := eventutil.RedactEvent(redactionEvent, eventToRedact); err != nil {
logrus.WithError(err).WithField("event_id", ev.EventID()).Error("Failed to redact event")
continue
}
Ristretto cache (#2563) * Try Ristretto cache * Tweak * It's beautiful * Update GMSL * More strict keyable interface * Fix that some more * Make less panicky * Don't enforce mutability checks for now * Determine mutability using deep equality * Tweaks * Namespace keys * Make federation caches mutable * Update cost estimation, add metric * Update GMSL * Estimate cost for metrics better * Reduce counters a bit * Try caching events * Some guards * Try again * Try this * Use separate caches for hopefully better hash distribution * Fix bug with admitting events into cache * Try to fix bugs * Check nil * Try that again * Preserve order jeezo this is messy * thanks VS Code for doing exactly the wrong thing * Try this again * Be more specific * aaaaargh * One more time * That might be better * Stronger sorting * Cache expiries, async publishing of EDUs * Put it back * Use a shared cache again * Cost estimation fixes * Update ristretto * Reduce counters a bit * Clean up a bit * Update GMSL * 1GB * Configurable cache sizees * Tweaks * Add `config.DataUnit` for specifying friendly cache sizes * Various tweaks * Update GMSL * Add back some lazy loading caching * Include key in cost * Include key in cost * Tweak max age handling, config key name * Only register prometheus metrics if requested * Review comments @S7evinK * Don't return errors when creating caches (it is better just to crash since otherwise we'll `nil`-pointer exception everywhere) * Review comments * Update sample configs * Update GHA Workflow * Update Complement images to Go 1.18 * Remove the cache test from the federation API as we no longer guarantee immediate cache admission * Don't check the caches in the renewal test * Possibly fix the upgrade tests * Update to matrix-org/gomatrixserverlib#322 * Update documentation to refer to Go 1.18
2022-07-11 15:31:31 +02:00
ev = eventToRedact.Headered(ev.RoomVersion)
events[j] = ev
}
backfilledEventMap[ev.EventID()] = types.Event{
EventNID: eventNID,
Event: ev.Unwrap(),
}
}
return roomNID, backfilledEventMap
}