[ https://issues.apache.org/jira/browse/FLINK-9377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16629470#comment-16629470 ]
ASF GitHub Bot commented on FLINK-9377: --------------------------------------- StephanEwen commented on a change in pull request #6711: [FLINK-9377] [core, state backends] Remove serializers from checkpoints URL: https://github.com/apache/flink/pull/6711#discussion_r220727617 ########## File path: flink-core/src/main/java/org/apache/flink/api/common/typeutils/TypeSerializerConfigSnapshotSerializationUtil.java ########## @@ -0,0 +1,172 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you 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 org.apache.flink.api.common.typeutils; + +import org.apache.flink.core.io.VersionedIOReadableWritable; +import org.apache.flink.core.memory.DataInputView; +import org.apache.flink.core.memory.DataOutputView; +import org.apache.flink.util.InstantiationUtil; +import org.apache.flink.util.Preconditions; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.List; + +/** + * Utility methods for serialization of {@link TypeSerializerConfigSnapshot}. + */ +public class TypeSerializerConfigSnapshotSerializationUtil { + + /** + * Writes a {@link TypeSerializerConfigSnapshot} to the provided data output view. + * + * <p>It is written with a format that can be later read again using + * {@link #readSerializerConfigSnapshot(DataInputView, ClassLoader)}. + * + * @param out the data output view + * @param serializerConfigSnapshot the serializer configuration snapshot to write + * + * @throws IOException + */ + public static <T> void writeSerializerConfigSnapshot( + DataOutputView out, + TypeSerializerConfigSnapshot<T> serializerConfigSnapshot, + TypeSerializer<T> serializer) throws IOException { + + new TypeSerializerConfigSnapshotSerializationProxy<>(serializerConfigSnapshot, serializer).write(out); + } + + /** + * Reads from a data input view a {@link TypeSerializerConfigSnapshot} that was previously + * written using {@link TypeSerializerConfigSnapshotSerializationUtil#writeSerializerConfigSnapshot(DataOutputView, TypeSerializerConfigSnapshot, TypeSerializer)}. + * + * @param in the data input view + * @param userCodeClassLoader the user code class loader to use + * + * @return the read serializer configuration snapshot + * + * @throws IOException + */ + public static TypeSerializerConfigSnapshot readSerializerConfigSnapshot( + DataInputView in, + ClassLoader userCodeClassLoader) throws IOException { + + final TypeSerializerConfigSnapshotSerializationProxy proxy = new TypeSerializerConfigSnapshotSerializationProxy(userCodeClassLoader); + proxy.read(in); + + return proxy.getSerializerConfigSnapshot(); + } + + /** + * Reads from a data input view multiple {@link TypeSerializerConfigSnapshot}s that was previously + * written using {@link TypeSerializerConfigSnapshotSerializationUtil#writeSerializerConfigSnapshot(DataOutputView, TypeSerializerConfigSnapshot, TypeSerializer)}. + * + * @param in the data input view + * @param userCodeClassLoader the user code class loader to use + * + * @return the read serializer configuration snapshots + * + * @throws IOException + */ + public static List<TypeSerializerConfigSnapshot<?>> readSerializerConfigSnapshots( + DataInputView in, + ClassLoader userCodeClassLoader) throws IOException { + + int numFields = in.readInt(); + final List<TypeSerializerConfigSnapshot<?>> serializerConfigSnapshots = new ArrayList<>(numFields); + + TypeSerializerConfigSnapshotSerializationProxy proxy; + for (int i = 0; i < numFields; i++) { + proxy = new TypeSerializerConfigSnapshotSerializationProxy(userCodeClassLoader); + proxy.read(in); + serializerConfigSnapshots.add(proxy.getSerializerConfigSnapshot()); + } + + return serializerConfigSnapshots; + } + + /** + * Utility serialization proxy for a {@link TypeSerializerConfigSnapshot}. + */ + static final class TypeSerializerConfigSnapshotSerializationProxy<T> extends VersionedIOReadableWritable { + + private static final int VERSION = 1; + + private ClassLoader userCodeClassLoader; + private TypeSerializerConfigSnapshot<T> serializerConfigSnapshot; + private TypeSerializer<T> serializer; + + TypeSerializerConfigSnapshotSerializationProxy(ClassLoader userCodeClassLoader) { + this.userCodeClassLoader = Preconditions.checkNotNull(userCodeClassLoader); + } + + TypeSerializerConfigSnapshotSerializationProxy( + TypeSerializerConfigSnapshot<T> serializerConfigSnapshot, + TypeSerializer<T> serializer) { + this.serializerConfigSnapshot = Preconditions.checkNotNull(serializerConfigSnapshot); + this.serializer = Preconditions.checkNotNull(serializer); + } + + @Override + public void write(DataOutputView out) throws IOException { + super.write(out); + + // config snapshot class, so that we can re-instantiate the + // correct type of config snapshot instance when deserializing + out.writeUTF(serializerConfigSnapshot.getClass().getName()); + + // TODO this is a temporary workaround until all serializer config snapshot classes in Flink + // TODO have properly implemented the restore serializer factory methods + serializerConfigSnapshot.setSerializer(serializer); + + // the actual configuration parameters + serializerConfigSnapshot.write(out); + } + + @SuppressWarnings("unchecked") + @Override + public void read(DataInputView in) throws IOException { + super.read(in); + + String serializerConfigClassname = in.readUTF(); + Class<? extends TypeSerializerConfigSnapshot> serializerConfigSnapshotClass; + try { + serializerConfigSnapshotClass = (Class<? extends TypeSerializerConfigSnapshot>) + Class.forName(serializerConfigClassname, true, userCodeClassLoader); + } catch (ClassNotFoundException e) { + throw new IOException( + "Could not find requested TypeSerializerConfigSnapshot class " + + serializerConfigClassname + " in classpath.", e); + } + + serializerConfigSnapshot = InstantiationUtil.instantiate(serializerConfigSnapshotClass); + serializerConfigSnapshot.setUserCodeClassLoader(userCodeClassLoader); Review comment: The separation between `setUserCodeClassLoader(userCodeClassLoader);` and `read(in)` seems a bit unclean, the methods always require to be used in conjunction. It also leads a field "classloader" in the snapshot which is sometimes set and sometimes not. That is typically an indication that this should be one method: `read(in, classloader)`. That does not match the `IOReadableWritable` interface, but I see not why the config snapshots need to have that specific interface. They could simply define their own interface. If this blows the scope of this change, I would suggest to plan a followup to clean this up. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > Remove writing serializers as part of the checkpoint meta information > --------------------------------------------------------------------- > > Key: FLINK-9377 > URL: https://issues.apache.org/jira/browse/FLINK-9377 > Project: Flink > Issue Type: Sub-task > Components: State Backends, Checkpointing > Reporter: Tzu-Li (Gordon) Tai > Assignee: Tzu-Li (Gordon) Tai > Priority: Critical > Labels: pull-request-available > Fix For: 1.7.0 > > > When writing meta information of a state in savepoints, we currently write > both the state serializer as well as the state serializer's configuration > snapshot. > Writing both is actually redundant, as most of the time they have identical > information. > Moreover, the fact that we use Java serialization to write the serializer > and rely on it to be re-readable on the restore run, already poses problems > for serializers such as the {{AvroSerializer}} (see discussion in FLINK-9202) > to perform even a compatible upgrade. > The proposal here is to leave only the config snapshot as meta information, > and use that as the single source of truth of information about the schema of > serialized state. > The config snapshot should be treated as a factory (or provided to a > factory) to re-create serializers capable of reading old, serialized state. -- This message was sent by Atlassian JIRA (v7.6.3#76005)