Github user tzulitai commented on a diff in the pull request: https://github.com/apache/flink/pull/3834#discussion_r115141541 --- Diff: flink-contrib/flink-statebackend-rocksdb/src/main/java/org/apache/flink/contrib/streaming/state/RocksDBKeyedStateBackend.java --- @@ -1473,22 +1481,92 @@ void restore(Collection<KeyedStateHandle> restoreStateHandles) throws Exception protected <N, S> ColumnFamilyHandle getColumnFamily( StateDescriptor<?, S> descriptor, TypeSerializer<N> namespaceSerializer) throws IOException { - Tuple2<ColumnFamilyHandle, RegisteredBackendStateMetaInfo<?, ?>> stateInfo = + Tuple2<ColumnFamilyHandle, RegisteredKeyedBackendStateMetaInfo<?, ?>> stateInfo = kvStateInformation.get(descriptor.getName()); - RegisteredBackendStateMetaInfo<N, S> newMetaInfo = new RegisteredBackendStateMetaInfo<>( - descriptor.getType(), - descriptor.getName(), - namespaceSerializer, - descriptor.getSerializer()); + RegisteredKeyedBackendStateMetaInfo<N, S> newMetaInfo = new RegisteredKeyedBackendStateMetaInfo<>( + descriptor.getType(), + descriptor.getName(), + namespaceSerializer, + descriptor.getSerializer()); if (stateInfo != null) { - if (newMetaInfo.canRestoreFrom(stateInfo.f1)) { + // TODO with eager registration in place, these checks should be moved to restore() + + RegisteredKeyedBackendStateMetaInfo.Snapshot<N, S> restoredMetaInfo = + restoredKvStateMetaInfos.get(descriptor.getName()); + + Preconditions.checkState( + newMetaInfo.getName().equals(restoredMetaInfo.getName()), + "Incompatible state names. " + + "Was [" + restoredMetaInfo.getName() + "], " + + "registered with [" + newMetaInfo.getName() + "]."); + + if (!newMetaInfo.getStateType().equals(StateDescriptor.Type.UNKNOWN) + && !restoredMetaInfo.getStateType().equals(StateDescriptor.Type.UNKNOWN)) { + + Preconditions.checkState( + newMetaInfo.getStateType().equals(restoredMetaInfo.getStateType()), + "Incompatible state types. " + + "Was [" + restoredMetaInfo.getStateType() + "], " + + "registered with [" + newMetaInfo.getStateType() + "]."); + } + + // check serializer migration strategies to determine if state migration is required + + boolean requireMigration = false; + + // only check migration strategy if there is a restored configuration snapshot; + // there wouldn't be one if we were restored from an older version checkpoint, + // in which case we can only simply assume that migration is not required + + if (restoredMetaInfo.getNamespaceSerializerConfigSnapshot() != null) { + MigrationStrategy<N> namespaceMigrationStrategy = newMetaInfo.getNamespaceSerializer() + .getMigrationStrategyFor(restoredMetaInfo.getNamespaceSerializerConfigSnapshot()); + + TypeSerializer<N> finalOldNamespaceSerializer; + if (namespaceMigrationStrategy.requireMigration()) { + requireMigration = true; + + if (namespaceMigrationStrategy.getFallbackDeserializer() != null) { + finalOldNamespaceSerializer = namespaceMigrationStrategy.getFallbackDeserializer(); + } else if (restoredMetaInfo.getNamespaceSerializer() != null + && !(restoredMetaInfo.getNamespaceSerializer() instanceof MigrationNamespaceSerializerProxy)) { + finalOldNamespaceSerializer = restoredMetaInfo.getNamespaceSerializer(); + } else { + throw new RuntimeException( + "State migration required, but there is no available serializer capable of reading previous namespace."); + } + } + } + + if (restoredMetaInfo.getStateSerializerConfigSnapshot() != null) { + MigrationStrategy<S> stateMigrationStrategy = newMetaInfo.getStateSerializer() + .getMigrationStrategyFor(restoredMetaInfo.getStateSerializerConfigSnapshot()); + + TypeSerializer<S> finalOldStateSerializer; + if (stateMigrationStrategy.requireMigration()) { + requireMigration = true; + + if (stateMigrationStrategy.getFallbackDeserializer() != null) { --- End diff -- You're correct with the setting that the fallback serializer has higher priority, in the current PR. But I agree that logically, the old Java serialized serializer should be a safer bet and should have higher priority. I think it would be a better balance to go with the flow that prioritizes the Java serialized serializer, but may just potentially create an unused fallback serializer. It is a "fallback", anyway, and with this flow the user implementation can be more compact (as you noticed, there a lot of duplicate logic for the user to implement if we split this up into 2 methods).
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---