fix a sneakybug

Author: ranlavanet

protocol/lavasession/consumer_session_manager.go

@@ -925,17 +925,6 @@ func (csm *ConsumerSessionManager) OnSessionFailure(consumerSession *SingleConsu
 		blockProvider = true
 	}
 
-	if sdkerrors.IsOf(errorReceived, BlockEndpointError) {
-		utils.LavaFormatTrace("Got BlockEndpointError, blocking endpoint and session",
-			utils.LogAttr("error", errorReceived),
-			utils.LogAttr("sessionID", consumerSession.SessionId),
-		)
-
-		// Block the endpoint and the consumer session from future usages
-		consumerSession.EndpointConnection.blockListed.Store(true)
-		consumerSession.BlockListed = true
-	}
-
 	consumerSession.QoSManager.AddFailedRelay(consumerSession.epoch, consumerSession.SessionId)
 	consumerSession.ConsecutiveErrors = append(consumerSession.ConsecutiveErrors, errorReceived)
 	// copy consecutive errors for report.

protocol/lavasession/consumer_session_manager_test.go

@@ -855,9 +855,12 @@ func TestAllProvidersEndpointsDisabled(t *testing.T) {
 	pairingList := createPairingList("", false)
 	err := csm.UpdateAllProviders(firstEpochHeight, pairingList) // update the providers.
 	require.NoError(t, err)
-	cs, err := csm.GetSessions(ctx, cuForFirstRequest, NewUsedProviders(nil), servicedBlockNumber, "", nil, common.NO_STATE, 0) // get a session
-	require.Nil(t, cs)
-	require.Error(t, err)
+	usedProviders := NewUsedProviders(nil)
+	cs, err := csm.GetSessions(ctx, cuForFirstRequest, usedProviders, servicedBlockNumber, "", nil, common.NO_STATE, 0) // get a session
+	require.NoError(t, err)
+	for key, _ := range cs {
+		require.Contains(t, csm.currentlyBlockedProviderAddresses, key)
+	}
 }
 
 func TestUpdateAllProviders(t *testing.T) {

protocol/lavasession/consumer_types.go

@@ -96,8 +96,10 @@ type EndpointConnection struct {
 	Client                              pairingtypes.RelayerClient
 	connection                          *grpc.ClientConn
 	numberOfSessionsUsingThisConnection uint64
-	blockListed                         atomic.Bool
-	lbUniqueId                          string
+	// blockListed - currently unused, use it carefully as it will block this provider's endpoint until next epoch without forgiveness.
+	// Can be used in cases of data reliability, self provider conflict etc..
+	blockListed atomic.Bool
+	lbUniqueId  string
 	// In case we got disconnected, we cant reconnect as we might lose stickiness
 	// with the provider, if its using a load balancer
 	disconnected bool
@@ -500,6 +502,7 @@ func (cswp *ConsumerSessionsWithProvider) fetchEndpointConnectionFromConsumerSes
 					if endpointConnection.Client != nil && endpointConnection.connection != nil && !endpointConnection.disconnected {
 						// Check if the endpoint is not blocked
 						if endpointConnection.blockListed.Load() {
+							utils.LavaFormatDebug("Skipping provider's endpoint as its block listed", utils.LogAttr("address", endpoint.NetworkAddress), utils.LogAttr("PublicLavaAddress", cswp.PublicLavaAddress))
 							continue
 						}
 						connectionState := endpointConnection.connection.GetState()

protocol/rpcconsumer/rpcconsumer_server.go

@@ -1014,7 +1014,7 @@ func (rpccs *RPCConsumerServer) relayInner(ctx context.Context, singleConsumerSe
 
 				if !singleConsumerSession.VerifyProviderUniqueIdAndStoreIfFirstTime(providerUniqueId[0]) {
 					return reply, 0, utils.LavaFormatError("provider unique id mismatch",
-						errors.Join(lavasession.SessionOutOfSyncError, lavasession.BlockEndpointError),
+						lavasession.SessionOutOfSyncError,
 						utils.LogAttr("GUID", ctx),
 						utils.LogAttr("sessionId", relayRequest.RelaySession.SessionId),
 						utils.LogAttr("provider", relayRequest.RelaySession.Provider),

scripts/init_chain.sh

@@ -59,7 +59,7 @@ else
         | jq '.app_state.downtime.params.downtime_duration = "6s"' \
         | jq '.app_state.downtime.params.epoch_duration = "10s"' \
         | jq '.app_state.epochstorage.params.epochsToSave = "8"' \
-        | jq '.app_state.epochstorage.params.epochBlocks = "20"' \
+        | jq '.app_state.epochstorage.params.epochBlocks = "240"' \
         | jq '.app_state.pairing.params.recommendedEpochNumToCollectPayment = "2"' \
     )
 fi