You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by kayousterhout <gi...@git.apache.org> on 2014/08/04 09:05:48 UTC
[GitHub] spark pull request: Remove support for waiting for executors in st...
GitHub user kayousterhout opened a pull request:
https://github.com/apache/spark/pull/1762
Remove support for waiting for executors in standalone mode.
Current code waits until some minimum fraction of expected executors
have registered before beginning scheduling. The current code in
standalone mode suffers from a race condition (SPARK-2635). This
race condition could be fixed, but this functionality is easily
achieved by the user (they can use the storage status to determine
how many executors are up, as described by @pwendell in #1462)
so adding the extra complexity to the scheduler code may not be worthwhile.
This commit removes the functionality in standalone mode but not for
YARN -- where it is more necessary and the number of expected executors
is well-defined.
This PR is a POC; if the powers-that-be determine that this is what we should
do, I will file a JIRA.
This should be backported into 1.1 if committed.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/kayousterhout/spark-1 remove_executor_wait
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/1762.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #1762
----
commit fa746ed65a8ac685edd33c79159521398d99aa69
Author: Kay Ousterhout <ka...@gmail.com>
Date: 2014-08-04T06:59:05Z
Remove support for waiting for executors in standalone mode.
Current code waits until some minimum fraction of expected executors
have registered before beginning scheduling. The current code in
standalone mode suffers from a race condition (SPARK-2635). This
race condition could be fixed, but this functionality is easily
achieved by the user (they can use the storage status to determine
how many executors are up, as described by @pwendell in #1462)
so adding the extra complexity to the scheduler code is not worthwile.
----
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [POC] Remove support for waiting for executors...
Posted by kayousterhout <gi...@git.apache.org>.
Github user kayousterhout closed the pull request at:
https://github.com/apache/spark/pull/1762
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: Remove support for waiting for executors in st...
Posted by kayousterhout <gi...@git.apache.org>.
Github user kayousterhout commented on a diff in the pull request:
https://github.com/apache/spark/pull/1762#discussion_r15742607
--- Diff: yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnSchedulerBackend.scala ---
@@ -0,0 +1,63 @@
+/*
+ * 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
+
+import akka.actor.ActorSystem
+
+import org.apache.spark.SparkContext
+import org.apache.spark.scheduler.TaskSchedulerImpl
+
+/**
+ * Subclass of CoarseGrainedSchedulerBackend that handles waiting until sufficient resources
+ * are registered before beginning to schedule tasks (needed by both Yarn scheduler backends).
+ */
+private[spark] class YarnSchedulerBackend(
+ scheduler: TaskSchedulerImpl,
+ actorSystem: ActorSystem)
+ extends CoarseGrainedSchedulerBackend(scheduler, actorSystem) {
+
+ // Submit tasks only after (registered executors / total expected executors)
+ // is equal to at least this value (expected to be a double between 0 and 1, inclusive).
+ var minRegisteredRatio = conf.getDouble("spark.scheduler.minRegisteredExecutorsRatio", 0.8)
+ if (minRegisteredRatio > 1) minRegisteredRatio = 1
+ // Regardless of whether the required number of executors have registered, return true from
+ // isReady() after this amount of time has elapsed.
+ val maxRegisteredWaitingTime =
+ conf.getInt("spark.scheduler.maxRegisteredExecutorsWaitingTime", 30000)
+ private val createTime = System.currentTimeMillis()
+
+ var totalExpectedExecutors: Int = _
+
+ override def isReady(): Boolean = {
--- End diff --
Realized I should fix this to only print the log message once -- will do this if we decide to use this PR
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org