From 3d8760d76eae41dcaab8e9aeda19619f3d5f1596 Mon Sep 17 00:00:00 2001
From: Andrew Or <andrew@databricks.com>
Date: Fri, 22 May 2015 17:37:38 -0700
Subject: [PATCH] [SPARK-7771] [SPARK-7779] Dynamic allocation: lower default
timeouts further
The default add time of 5s is still too slow for small jobs. Also, the current default remove time of 10 minutes seem rather high. This patch lowers both and rephrases a few log messages.
Author: Andrew Or <andrew@databricks.com>
Closes #6301 from andrewor14/da-minor and squashes the following commits:
6d614a6 [Andrew Or] Lower log level
2811492 [Andrew Or] Log information when requests are canceled
5fcd3eb [Andrew Or] Fix tests
3320710 [Andrew Or] Lower timeouts + rephrase a few log messages
---
.../spark/ExecutorAllocationManager.scala | 26 +++++++++++++------
docs/configuration.md | 4 +--
2 files changed, 20 insertions(+), 10 deletions(-)
diff --git a/core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala b/core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala
index 66bda68088..9514604752 100644
--- a/core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala
+++ b/core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala
@@ -91,7 +91,7 @@ private[spark] class ExecutorAllocationManager(
// How long there must be backlogged tasks for before an addition is triggered (seconds)
private val schedulerBacklogTimeoutS = conf.getTimeAsSeconds(
- "spark.dynamicAllocation.schedulerBacklogTimeout", "5s")
+ "spark.dynamicAllocation.schedulerBacklogTimeout", "1s")
// Same as above, but used only after `schedulerBacklogTimeoutS` is exceeded
private val sustainedSchedulerBacklogTimeoutS = conf.getTimeAsSeconds(
@@ -99,7 +99,7 @@ private[spark] class ExecutorAllocationManager(
// How long an executor must be idle for before it is removed (seconds)
private val executorIdleTimeoutS = conf.getTimeAsSeconds(
- "spark.dynamicAllocation.executorIdleTimeout", "600s")
+ "spark.dynamicAllocation.executorIdleTimeout", "60s")
// During testing, the methods to actually kill and add executors are mocked out
private val testing = conf.getBoolean("spark.dynamicAllocation.testing", false)
@@ -268,6 +268,8 @@ private[spark] class ExecutorAllocationManager(
numExecutorsTarget = math.max(maxNeeded, minNumExecutors)
client.requestTotalExecutors(numExecutorsTarget)
numExecutorsToAdd = 1
+ logInfo(s"Lowering target number of executors to $numExecutorsTarget because " +
+ s"not all requests are actually needed (previously $oldNumExecutorsTarget)")
numExecutorsTarget - oldNumExecutorsTarget
} else if (addTime != NOT_SET && now >= addTime) {
val delta = addExecutors(maxNeeded)
@@ -292,9 +294,8 @@ private[spark] class ExecutorAllocationManager(
private def addExecutors(maxNumExecutorsNeeded: Int): Int = {
// Do not request more executors if it would put our target over the upper bound
if (numExecutorsTarget >= maxNumExecutors) {
- val numExecutorsPending = numExecutorsTarget - executorIds.size
- logDebug(s"Not adding executors because there are already ${executorIds.size} registered " +
- s"and ${numExecutorsPending} pending executor(s) (limit $maxNumExecutors)")
+ logDebug(s"Not adding executors because our current target total " +
+ s"is already $numExecutorsTarget (limit $maxNumExecutors)")
numExecutorsToAdd = 1
return 0
}
@@ -310,10 +311,19 @@ private[spark] class ExecutorAllocationManager(
// Ensure that our target fits within configured bounds:
numExecutorsTarget = math.max(math.min(numExecutorsTarget, maxNumExecutors), minNumExecutors)
+ val delta = numExecutorsTarget - oldNumExecutorsTarget
+
+ // If our target has not changed, do not send a message
+ // to the cluster manager and reset our exponential growth
+ if (delta == 0) {
+ numExecutorsToAdd = 1
+ return 0
+ }
+
val addRequestAcknowledged = testing || client.requestTotalExecutors(numExecutorsTarget)
if (addRequestAcknowledged) {
- val delta = numExecutorsTarget - oldNumExecutorsTarget
- logInfo(s"Requesting $delta new executor(s) because tasks are backlogged" +
+ val executorsString = "executor" + { if (delta > 1) "s" else "" }
+ logInfo(s"Requesting $delta new $executorsString because tasks are backlogged" +
s" (new desired total will be $numExecutorsTarget)")
numExecutorsToAdd = if (delta == numExecutorsToAdd) {
numExecutorsToAdd * 2
@@ -420,7 +430,7 @@ private[spark] class ExecutorAllocationManager(
* This resets all variables used for adding executors.
*/
private def onSchedulerQueueEmpty(): Unit = synchronized {
- logDebug(s"Clearing timer to add executors because there are no more pending tasks")
+ logDebug("Clearing timer to add executors because there are no more pending tasks")
addTime = NOT_SET
numExecutorsToAdd = 1
}
diff --git a/docs/configuration.md b/docs/configuration.md
index 0de824546c..30508a617f 100644
--- a/docs/configuration.md
+++ b/docs/configuration.md
@@ -1194,7 +1194,7 @@ Apart from these, the following properties are also available, and may be useful
</tr>
<tr>
<td><code>spark.dynamicAllocation.executorIdleTimeout</code></td>
- <td>600s</td>
+ <td>60s</td>
<td>
If dynamic allocation is enabled and an executor has been idle for more than this duration,
the executor will be removed. For more detail, see this
@@ -1224,7 +1224,7 @@ Apart from these, the following properties are also available, and may be useful
</tr>
<tr>
<td><code>spark.dynamicAllocation.schedulerBacklogTimeout</code></td>
- <td>5s</td>
+ <td>1s</td>
<td>
If dynamic allocation is enabled and there have been pending tasks backlogged for more than
this duration, new executors will be requested. For more detail, see this
--
GitLab