acrites commented on code in PR #34053: URL: https://github.com/apache/beam/pull/34053#discussion_r1968707243
########## runners/google-cloud-dataflow-java/worker/src/main/java/org/apache/beam/runners/dataflow/worker/streaming/harness/BackendWorkerMetadataVendor.java: ########## @@ -0,0 +1,118 @@ +/* + * 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.beam.runners.dataflow.worker.streaming.harness; + +import com.google.common.util.concurrent.ThreadFactoryBuilder; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; +import java.util.concurrent.TimeUnit; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.function.Consumer; +import java.util.function.Function; +import javax.annotation.concurrent.ThreadSafe; +import org.apache.beam.runners.dataflow.worker.windmill.WindmillEndpoints; +import org.apache.beam.runners.dataflow.worker.windmill.client.WindmillStream.GetWorkerMetadataStream; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +@ThreadSafe +final class BackendWorkerMetadataVendor { + private static final Logger LOG = LoggerFactory.getLogger(BackendWorkerMetadataVendor.class); + private static final int GET_WORKER_METADATA_STREAM_TIMEOUT_MINUTES = 3; + private static final String BACKEND_WORKER_METADATA_VENDOR_THREAD_NAME = + "BackendWorkerMetadataVendorThread"; + + private final Function<Consumer<WindmillEndpoints>, GetWorkerMetadataStream> + getWorkerMetadataStreamFactory; + private final ExecutorService workerMetadataFetcher; + private final AtomicBoolean isRunning = new AtomicBoolean(false); + + private BackendWorkerMetadataVendor( + Function<Consumer<WindmillEndpoints>, GetWorkerMetadataStream> getWorkerMetadataStreamFactory, + ExecutorService workerMetadataFetcher) { + this.getWorkerMetadataStreamFactory = getWorkerMetadataStreamFactory; + this.workerMetadataFetcher = workerMetadataFetcher; + } + + static BackendWorkerMetadataVendor create( + Function<Consumer<WindmillEndpoints>, GetWorkerMetadataStream> + getWorkerMetadataStreamFactory) { + return new BackendWorkerMetadataVendor( + getWorkerMetadataStreamFactory, + Executors.newSingleThreadExecutor( + new ThreadFactoryBuilder() + .setNameFormat(BACKEND_WORKER_METADATA_VENDOR_THREAD_NAME) + .build())); + } + + /** + * Starts the vending of backend worker metadata, passing it to endpointsConsumer for processing. + */ + void start(Consumer<WindmillEndpoints> endpointsConsumer) { + if (isRunning.compareAndSet(false, true)) { + LOG.debug("Starting WorkerMetadataVendor..."); + // Block on the initial workerMetadata. + // Run the subsequent fetches on a separate thread. + fetchAndConsumeWorkerMetadata(endpointsConsumer); + LOG.debug("Initial WorkerMetadata received."); + workerMetadataFetcher.execute( + () -> { + while (isRunning.get()) { + fetchAndConsumeWorkerMetadata(endpointsConsumer); + } + }); + } + } + + void stop() { + if (isRunning.compareAndSet(true, false)) { + LOG.debug("Shutting down WorkerMetadataVendor..."); + workerMetadataFetcher.shutdown(); + boolean isShutdown = false; + try { + isShutdown = + workerMetadataFetcher.awaitTermination( + GET_WORKER_METADATA_STREAM_TIMEOUT_MINUTES, TimeUnit.MINUTES); + } catch (InterruptedException e) { + Thread.currentThread().interrupt(); + LOG.warn("Unable to shutdown WorkerMetadataFetcher."); + } + + if (!isShutdown) { + workerMetadataFetcher.shutdownNow(); + } + } + } + + private void fetchAndConsumeWorkerMetadata(Consumer<WindmillEndpoints> endpointsConsumer) { + GetWorkerMetadataStream getWorkerMetadataStream = + getWorkerMetadataStreamFactory.apply(endpointsConsumer); + try { + getWorkerMetadataStream.start(); + // Reconnect every now and again to enable better load balancing and to avoid hitting + // gRPC deadlines. If at any point the server closes the stream, we will reconnect + // immediately; otherwise shutdown the stream after some time and create a new one. + if (!getWorkerMetadataStream.awaitTermination( + GET_WORKER_METADATA_STREAM_TIMEOUT_MINUTES, TimeUnit.MINUTES)) { + getWorkerMetadataStream.shutdown(); Review Comment: I don't quite know what our protocol is here, but I thought we wanted to do something like half-close on the client so that the WM worker can flush all remaining work and then close on its end? If that's the case it seems like we'd want to half-close before calling `awaitTermination` (with a smaller timeout?). We'd need to have some other sleep instead (for up to 3 minutes maybe) before we called halfClose. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: github-unsubscr...@beam.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
