[GitHub] spark pull request #21366: [SPARK-24248][K8S] Use the Kubernetes API to popu...

[mccheah <gi...@git.apache.org>]

Github user mccheah commented on a diff in the pull request:

    https://github.com/apache/spark/pull/21366#discussion_r192271761
  
    --- Diff: resource-managers/kubernetes/core/src/main/scala/org/apache/spark/scheduler/cluster/k8s/ExecutorPodsLifecycleEventHandler.scala ---
    @@ -0,0 +1,130 @@
    +/*
    + * 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.scheduler.cluster.k8s
    +
    +import com.google.common.cache.{Cache, CacheBuilder}
    +import io.fabric8.kubernetes.api.model.Pod
    +import io.fabric8.kubernetes.client.KubernetesClient
    +import scala.collection.JavaConverters._
    +
    +import org.apache.spark.SparkConf
    +import org.apache.spark.deploy.k8s.Config._
    +import org.apache.spark.deploy.k8s.Constants._
    +import org.apache.spark.scheduler.ExecutorExited
    +import org.apache.spark.util.Utils
    +
    +private[spark] class ExecutorPodsLifecycleEventHandler(
    +    conf: SparkConf,
    +    executorBuilder: KubernetesExecutorBuilder,
    +    kubernetesClient: KubernetesClient,
    +    podsEventQueue: ExecutorPodsEventQueue,
    +    // Use a best-effort to track which executors have been removed already. It's not generally
    +    // job-breaking if we remove executors more than once but it's ideal if we make an attempt
    +    // to avoid doing so. Expire cache entries so that this data structure doesn't grow beyond
    +    // bounds.
    +    removedExecutorsCache: Cache[java.lang.Long, java.lang.Long]) {
    +
    +  import ExecutorPodsLifecycleEventHandler._
    +
    +  private val eventProcessingInterval = conf.get(KUBERNETES_EXECUTOR_EVENT_PROCESSING_INTERVAL)
    +
    +  def start(schedulerBackend: KubernetesClusterSchedulerBackend): Unit = {
    +    podsEventQueue.addSubscriber(eventProcessingInterval) { updatedPods =>
    +      updatedPods.foreach { updatedPod =>
    +        processUpdatedPod(schedulerBackend, updatedPod)
    +      }
    +    }
    +  }
    +
    +  private def processUpdatedPod(
    +      schedulerBackend: KubernetesClusterSchedulerBackend, updatedPod: Pod) = {
    +    val execId = updatedPod.getMetadata.getLabels.get(SPARK_EXECUTOR_ID_LABEL).toLong
    +    if (isDeleted(updatedPod)) {
    +      removeExecutorFromSpark(schedulerBackend, updatedPod, execId)
    --- End diff --
    
    It's not entirely analogous but there are parts that can be shared. The pods allocator has a "post-batch" processing task which can run even on empty batches (particularly important for the first empty batch to request the first round of executors). But, I have some ideas for shared code that can be posted shortly.


---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org