Github user steveloughran commented on a diff in the pull request: https://github.com/apache/spark/pull/8744#discussion_r43387678 --- Diff: yarn/src/history/main/scala/org/apache/spark/deploy/history/yarn/YarnHistoryService.scala --- @@ -0,0 +1,1328 @@ +/* + * 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.spark.deploy.history.yarn + +import java.io.InterruptedIOException +import java.net.{ConnectException, URI} +import java.util.concurrent.atomic.{AtomicBoolean, AtomicInteger} +import java.util.concurrent.{LinkedBlockingDeque, TimeUnit} + +import scala.collection.JavaConverters._ +import scala.collection.mutable + +import com.codahale.metrics.{Counter, MetricRegistry, Timer} +import org.apache.hadoop.security.UserGroupInformation +import org.apache.hadoop.yarn.api.records.timeline.{TimelineDomain, TimelineEntity, TimelineEvent} +import org.apache.hadoop.yarn.api.records.{ApplicationAttemptId, ApplicationId} +import org.apache.hadoop.yarn.client.api.TimelineClient +import org.apache.hadoop.yarn.conf.YarnConfiguration + +import org.apache.spark.deploy.history.yarn.YarnTimelineUtils._ +import org.apache.spark.metrics.source.Source +import org.apache.spark.scheduler._ +import org.apache.spark.scheduler.cluster.{SchedulerExtensionService, SchedulerExtensionServiceBinding} +import org.apache.spark.util.{SystemClock, Utils} +import org.apache.spark.{Logging, SparkContext} + +/** + * A Yarn Extension Service to post lifecycle events to a registered YARN Timeline Server. + * + * Posting algorithm + * + * 1. The service subscribes to all events coming from the Spark Context. + * 1. These events are serialized into JSON objects for publishing to the timeline service through + * HTTP(S) posts. + * 1. Events are buffered into `pendingEvents` until a batch is aggregated into a + * [[TimelineEntity]] for posting. + * 1. That aggregation happens when a lifecycle event (application start/stop) takes place, + * or the number of pending events in a running application exceeds the limit set in + * `spark.hadoop.yarn.timeline.batch.size`. + * 1. Posting operations take place in a separate thread from the spark event listener. + * 1. If an attempt to post to the timeline server fails, the service sleeps and then + * it is re-attempted after the retry period defined by + * `spark.hadoop.yarn.timeline.post.retry.interval`. + * 1. If the number of events buffered in the history service exceed the limit set in + * `spark.hadoop.yarn.timeline.post.limit`, then further events other than application start/stop + * are dropped. + * 1. When the service is stopped, it will make a best-effort attempt to post all queued events. + * the call of [[stop()]] can block up to the duration of + * `spark.hadoop.yarn.timeline.shutdown.waittime` for this to take place. + * 1. No events are posted until the service receives a [[SparkListenerApplicationStart]] event. + * + * If the spark context has a metrics registry, then the internal counters of queued entities, + * post failures and successes, and the performance of the posting operation are all registered + * as metrics. + * + * The shutdown logic is somewhat convoluted, as the posting thread may be blocked on HTTP IO + * when the shutdown process begins. In this situation, the thread continues to be blocked, and + * will be interrupted once the wait time has expired. All time consumed during the ongoing + * operation will be counted as part of the shutdown time period. + */ +private[spark] class YarnHistoryService extends SchedulerExtensionService + with Logging with Source { --- End diff -- OK, I'll pull out the counters into their source subclass
--- 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. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org