Try to fix backfilling (#2548)

* Try to fix backfilling

* Return start/end to not confuse clients

* Update GMSL

* Update GMSL

go.mod

@@ -34,7 +34,7 @@ require (
 	github.com/matrix-org/dugong v0.0.0-20210921133753-66e6b1c67e2e
 	github.com/matrix-org/go-sqlite3-js v0.0.0-20220419092513-28aa791a1c91
 	github.com/matrix-org/gomatrix v0.0.0-20210324163249-be2af5ef2e16
-	github.com/matrix-org/gomatrixserverlib v0.0.0-20220613132209-aedb3fbb511a
+	github.com/matrix-org/gomatrixserverlib v0.0.0-20220701090733-da53994b0c7f
 	github.com/matrix-org/pinecone v0.0.0-20220408153826-2999ea29ed48
 	github.com/matrix-org/util v0.0.0-20200807132607-55161520e1d4
 	github.com/mattn/go-sqlite3 v1.14.13

go.sum

@@ -418,8 +418,8 @@ github.com/matrix-org/go-sqlite3-js v0.0.0-20220419092513-28aa791a1c91/go.mod h1
 github.com/matrix-org/gomatrix v0.0.0-20190528120928-7df988a63f26/go.mod h1:3fxX6gUjWyI/2Bt7J1OLhpCzOfO/bB3AiX0cJtEKud0=
 github.com/matrix-org/gomatrix v0.0.0-20210324163249-be2af5ef2e16 h1:ZtO5uywdd5dLDCud4r0r55eP4j9FuUNpl60Gmntcop4=
 github.com/matrix-org/gomatrix v0.0.0-20210324163249-be2af5ef2e16/go.mod h1:/gBX06Kw0exX1HrwmoBibFA98yBk/jxKpGVeyQbff+s=
-github.com/matrix-org/gomatrixserverlib v0.0.0-20220613132209-aedb3fbb511a h1:jOkrb6twViAGTHHadA51sQwdloHT0Vx1MCptk9InTHo=
+github.com/matrix-org/gomatrixserverlib v0.0.0-20220701090733-da53994b0c7f h1:XF2+J6sOq07yhK1I7ItwsgRwXorjj7gqiCvgZ4dn8W8=
-github.com/matrix-org/gomatrixserverlib v0.0.0-20220613132209-aedb3fbb511a/go.mod h1:jX38yp3SSLJNftBg3PXU1ayd0PCLIiDHQ4xAc9DIixk=
+github.com/matrix-org/gomatrixserverlib v0.0.0-20220701090733-da53994b0c7f/go.mod h1:jX38yp3SSLJNftBg3PXU1ayd0PCLIiDHQ4xAc9DIixk=
 github.com/matrix-org/pinecone v0.0.0-20220408153826-2999ea29ed48 h1:W0sjjC6yjskHX4mb0nk3p0fXAlbU5bAFUFeEtlrPASE=
 github.com/matrix-org/pinecone v0.0.0-20220408153826-2999ea29ed48/go.mod h1:ulJzsVOTssIVp1j/m5eI//4VpAGDkMt5NrRuAVX7wpc=
 github.com/matrix-org/util v0.0.0-20190711121626-527ce5ddefc7/go.mod h1:vVQlW/emklohkZnOPwD3LrZUBqdfsbiyO3p1lNV8F6U=
@@ -482,8 +482,6 @@ github.com/neelance/astrewrite v0.0.0-20160511093645-99348263ae86/go.mod h1:kHJE
 github.com/neelance/sourcemap v0.0.0-20151028013722-8c68805598ab/go.mod h1:Qr6/a/Q4r9LP1IltGz7tA7iOK1WonHEYhu1HRBA7ZiM=
 github.com/neilalexander/nats-server/v2 v2.8.3-0.20220513095553-73a9a246d34f h1:Fc+TjdV1mOy0oISSzfoxNWdTqjg7tN/Vdgf+B2cwvdo=
 github.com/neilalexander/nats-server/v2 v2.8.3-0.20220513095553-73a9a246d34f/go.mod h1:vIdpKz3OG+DCg4q/xVPdXHoztEyKDWRtykQ4N7hd7C4=
-github.com/neilalexander/nats.go v1.13.1-0.20220419101051-b262d9f0be1e h1:kNIzIzj2OvnlreA+sTJ12nWJzTP3OSLNKDL/Iq9mF6Y=
-github.com/neilalexander/nats.go v1.13.1-0.20220419101051-b262d9f0be1e/go.mod h1:BPko4oXsySz4aSWeFgOHLZs3G4Jq4ZAyE6/zMCxRT6w=
 github.com/neilalexander/nats.go v1.13.1-0.20220621084451-ac518c356673 h1:TcKfa3Tf0qwUotv63PQVu2d1bBoLi2iEA4RHVMGDh5M=
 github.com/neilalexander/nats.go v1.13.1-0.20220621084451-ac518c356673/go.mod h1:BPko4oXsySz4aSWeFgOHLZs3G4Jq4ZAyE6/zMCxRT6w=
 github.com/neilalexander/utp v0.1.1-0.20210727203401-54ae7b1cd5f9 h1:lrVQzBtkeQEGGYUHwSX1XPe1E5GL6U3KYCNe2G4bncQ=

roomserver/internal/perform/perform_backfill.go

@@ -18,6 +18,7 @@ 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"
@@ -206,8 +207,17 @@ func (r *Backfiller) fetchAndStoreMissingEvents(ctx context.Context, roomVer gom
 			}
 			logger.Infof("returned %d PDUs which made events %+v", len(res.PDUs), result)
 			for _, res := range result {
-				if res.Error != nil {
+				switch err := res.Error.(type) {
-					logger.WithError(res.Error).Warn("event failed PDU checks")
+				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
@@ -306,6 +316,7 @@ FederationHit:
 		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
 }
 
@@ -366,20 +377,26 @@ func (b *backfillRequester) StateBeforeEvent(ctx context.Context, roomVer gomatr
 		}
 	}
 
+	var lastErr error
+	for _, srv := range b.servers {
 		c := gomatrixserverlib.FederatedStateProvider{
 			FedClient:          b.fsAPI,
 			RememberAuthEvents: false,
-		Server:             b.servers[0],
+			Server:             srv,
 		}
 		result, err := c.StateBeforeEvent(ctx, roomVer, event, eventIDs)
 		if err != nil {
-		return nil, err
+			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

syncapi/routing/context.go

@@ -15,6 +15,7 @@
 package routing
 
 import (
+	"context"
 	"database/sql"
 	"encoding/json"
 	"fmt"
@@ -25,6 +26,7 @@ import (
 	"github.com/matrix-org/dendrite/internal/caching"
 	roomserver "github.com/matrix-org/dendrite/roomserver/api"
 	"github.com/matrix-org/dendrite/syncapi/storage"
+	"github.com/matrix-org/dendrite/syncapi/types"
 	userapi "github.com/matrix-org/dendrite/userapi/api"
 	"github.com/matrix-org/gomatrixserverlib"
 	"github.com/matrix-org/util"
@@ -149,13 +151,30 @@ func Context(
 	if len(response.State) > filter.Limit {
 		response.State = response.State[len(response.State)-filter.Limit:]
 	}
-
+	start, end, err := getStartEnd(ctx, syncDB, eventsBefore, eventsAfter)
+	if err == nil {
+		response.End = end.String()
+		response.Start = start.String()
+	}
 	return util.JSONResponse{
 		Code: http.StatusOK,
 		JSON: response,
 	}
 }
 
+func getStartEnd(ctx context.Context, syncDB storage.Database, startEvents, endEvents []*gomatrixserverlib.HeaderedEvent) (start, end types.TopologyToken, err error) {
+	if len(startEvents) > 0 {
+		start, err = syncDB.EventPositionInTopology(ctx, startEvents[0].EventID())
+		if err != nil {
+			return
+		}
+	}
+	if len(endEvents) > 0 {
+		end, err = syncDB.EventPositionInTopology(ctx, endEvents[0].EventID())
+	}
+	return
+}
+
 func applyLazyLoadMembers(
 	device *userapi.Device,
 	filter *gomatrixserverlib.RoomEventFilter,