## What changes were proposed in this pull request?
When a fatal error (such as StackOverflowError) throws from "receiveAndReply", we should try our best to notify the sender. Otherwise, the sender will hang until timeout.
In addition, when a MessageLoop is dying unexpectedly, it should resubmit a new one so that Dispatcher is still working.
## How was this patch tested?
New unit tests.
Closes#24396 from zsxwing/SPARK-27496.
Authored-by: Shixiong Zhu <zsxwing@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
We have disabled a test related to storage in the History server suite after SPARK-13845. But, after SPARK-22050, we can store the information about block updated events to eventLog, if we enable "spark.eventLog.logBlockUpdates.enabled=true".
So, we can enable the test, by adding an eventlog corresponding to the application, which has enabled the configuration, "spark.eventLog.logBlockUpdates.enabled=true"
## How was this patch tested?
Existing UTs
Closes#24390 from shahidki31/enableRddStorageTest.
Authored-by: Shahid <shahidki31@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Finish the rest work of https://github.com/apache/spark/pull/24317, https://github.com/apache/spark/pull/9030
a. Implement Kryo serialization for UnsafeArrayData
b. fix UnsafeMapData Java/Kryo Serialization issue when two machines have different Oops size
c. Move the duplicate code "getBytes()" to Utils.
## How was this patch tested?
According Units has been added & tested
Closes#24357 from pengbo/SPARK-27416_new.
Authored-by: pengbo <bo.peng1019@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Added Constant instead of referring the same String literal "spark.buffer.pageSize" from many places
## How was this patch tested?
Run the corresponding Unit Test Cases manually.
Closes#24368 from shivusondur/Constant.
Authored-by: shivusondur <shivusondur@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
This PR supports `OpenJ9` in addition to `IBM JDK` and `OpenJDK` in Spark by handling `System.getProperty("java.vendor") = "Eclipse OpenJ9"`.
In `inferDefaultMemory()` and `getKrb5LoginModuleName()`, this PR uses non `IBM` way.
```
$ ~/jdk-11.0.2+9_openj9-0.12.1/bin/jshell
| Welcome to JShell -- Version 11.0.2
| For an introduction type: /help intro
jshell> System.out.println(System.getProperty("java.vendor"))
Eclipse OpenJ9
jshell> System.out.println(System.getProperty("java.vm.info"))
JRE 11 Linux amd64-64-Bit Compressed References 20190204_127 (JIT enabled, AOT enabled)
OpenJ9 - 90dd8cb40
OMR - d2f4534b
JCL - 289c70b6844 based on jdk-11.0.2+9
jshell> System.out.println(Class.forName("com.ibm.lang.management.OperatingSystemMXBean").getDeclaredMethod("getTotalPhysicalMemory"))
public abstract long com.ibm.lang.management.OperatingSystemMXBean.getTotalPhysicalMemory()
jshell> System.out.println(Class.forName("com.sun.management.OperatingSystemMXBean").getDeclaredMethod("getTotalPhysicalMemorySize"))
public abstract long com.sun.management.OperatingSystemMXBean.getTotalPhysicalMemorySize()
jshell> System.out.println(Class.forName("com.ibm.security.auth.module.Krb5LoginModule"))
| Exception java.lang.ClassNotFoundException: com.ibm.security.auth.module.Krb5LoginModule
| at Class.forNameImpl (Native Method)
| at Class.forName (Class.java:339)
| at (#1:1)
jshell> System.out.println(Class.forName("com.sun.security.auth.module.Krb5LoginModule"))
class com.sun.security.auth.module.Krb5LoginModule
```
## How was this patch tested?
Existing test suites
Manual testing with OpenJ9.
Closes#24308 from kiszk/SPARK-27397.
Authored-by: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Update pyrolite to 4.23 to pick up bug and security fixes.
## How was this patch tested?
Existing tests.
Closes#24381 from srowen/SPARK-27470.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Our customer has a very complicated job. Sometimes it successes and sometimes it fails with
```
Caused by: org.apache.spark.SparkException: Job aborted due to stage failure: ShuffleMapStage 4 has failed the maximum allowable number of times: 4.
Most recent failure reason: org.apache.spark.shuffle.FetchFailedException
```
However, with the patch https://github.com/apache/spark/pull/23871 , the job hangs forever.
When I investigated it, I found that `DAGScheduler` and `TaskSchedulerImpl` define stage completion differently. `DAGScheduler` thinks a stage is completed if all its partitions are marked as completed ([result stage](https://github.com/apache/spark/blob/v2.4.1/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala#L1362-L1368) and [shuffle stage](https://github.com/apache/spark/blob/v2.4.1/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala#L1400)). `TaskSchedulerImpl` thinks a stage's task set is completed when all tasks finish (see the [code](https://github.com/apache/spark/blob/v2.4.1/core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala#L779-L784)).
Ideally this two definition should be consistent, but #23871 breaks it. In our customer's Spark log, I found that, a stage's task set completes, but the stage never completes. More specifically, `DAGScheduler` submits a task set for stage 4.1 with 1000 tasks, but the `TaskSetManager` skips to run the first 100 tasks. Later on, `TaskSetManager` finishes 900 tasks and marks the task set as completed. However, `DAGScheduler` doesn't agree with it and hangs forever, waiting for more task completion events of stage 4.1.
With hindsight, I think `TaskSchedulerIImpl.stageIdToFinishedPartitions` is fragile. We need to pay more effort to make sure this is consistent with `DAGScheduler`'s knowledge. When `DAGScheduler` marks some partitions from finished to unfinished, `TaskSchedulerIImpl.stageIdToFinishedPartitions` should be updated as well.
This PR reverts #23871, let's think of a more robust idea later.
## How was this patch tested?
N/A
Closes#24359 from cloud-fan/revert.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
The RBackend and RBackendHandler create new conf objects that don't pick up conf values from the existing SparkSession and therefore always use the default conf values instead of values specified by the user.
In this fix we check to see if the spark env already exists, and get the conf from there. We fall back to creating a new conf. This follows the pattern used in other places including this: 3725b1324f/core/src/main/scala/org/apache/spark/api/r/BaseRRunner.scala (L261)
## How was this patch tested?
(Please explain how this patch was tested. E.g. unit tests, integration tests, manual tests)
(If this patch involves UI changes, please attach a screenshot; otherwise, remove this)
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#24353 from MrBago/r-backend-use-existing-conf.
Authored-by: Bago Amirbekian <bago@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Fix build warnings -- see some details below.
But mostly, remove use of postfix syntax where it causes warnings without the `scala.language.postfixOps` import. This is mostly in expressions like "120000 milliseconds". Which, I'd like to simplify to things like "2.minutes" anyway.
## How was this patch tested?
Existing tests.
Closes#24314 from srowen/SPARK-27404.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
I propose to disable (comment) result checking in `SorterSuite`.`java.lang.ArrayIndexOutOfBoundsException in TimSort` because:
1. The check is optional, and correctness of TimSort is checked by another tests. Purpose of the test is to check that TimSort doesn't fail with `ArrayIndexOutOfBoundsException`.
2. Significantly drops execution time of the test.
Here are timing of running the test locally:
```
Sort: 1.4 seconds
Result checking: 15.6 seconds
```
## How was this patch tested?
By `SorterSuite`.
Closes#24343 from MaxGekk/timsort-test-speedup.
Authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
This PR updates `AppStatusListener` to flush `LiveEntity` if necessary when receiving `SparkListenerExecutorMetricsUpdate`. This will ensure the staleness of Spark UI doesn't last more than the executor heartbeat interval.
## How was this patch tested?
The new unit test.
Closes#24303 from zsxwing/SPARK-27394.
Authored-by: Shixiong Zhu <zsxwing@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
When you submit a stage on a large cluster, rack resolving takes a long time when initializing TaskSetManager because a script is invoked to resolve the rack of each host, one by one.
Based on current implementation, it takes 30~40 seconds to resolve the racks in our 5000 nodes' cluster. After applied the patch, it decreased to less than 15 seconds.
YARN-9332 has added an interface to handle multiple hosts in one invocation to save time. But before upgrading to the newest Hadoop, we could construct the same tool in Spark to resolve this issue.
## How was this patch tested?
UT and manually testing on a 5000 node cluster.
Closes#24245 from squito/SPARK-13704_update.
Lead-authored-by: LantaoJin <jinlantao@gmail.com>
Co-authored-by: Imran Rashid <irashid@cloudera.com>
Signed-off-by: Imran Rashid <irashid@cloudera.com>
## What changes were proposed in this pull request?
check spark.task.cpus before creating TaskScheduler in SparkContext
## How was this patch tested?
UT
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#24261 from liutang123/SPARK-27192.
Authored-by: liulijia <liutang123@yeah.net>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
This PR aims to clean up package name mismatches.
## How was this patch tested?
Pass the Jenkins.
Closes#24300 from dongjoon-hyun/SPARK-27390.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Update jquery -> 1.12.4, datatables -> 1.10.18, mustache -> 2.3.12.
Add missing mustache license
## How was this patch tested?
I manually tested the UI locally with the javascript console open and didn't observe any problems or JS errors. The only 'risky' change seems to be mustache, but on reading its release notes, don't think the changes from 0.8.1 to 2.x would affect Spark's simple usage.
Closes#24288 from srowen/SPARK-27358.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
https://github.com/apache/spark/pull/24265 breaks the lint check, because it has trailing space. (not sure why it passed jenkins). This PR fixes it.
## How was this patch tested?
N/A
Closes#24289 from cloud-fan/fix.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
In `UnsafeExternalSorter.SpillableIterator#loadNext()` takes lock on the `UnsafeExternalSorter` and calls `freePage` once the `lastPage` is consumed which needs to take a lock on `TaskMemoryManager`. At the same time, there can be another MemoryConsumer using `UnsafeExternalSorter` as part of sorting can try to `allocatePage` needs to get lock on `TaskMemoryManager` which can cause spill to happen which requires lock on `UnsafeExternalSorter` again causing deadlock. This is a classic deadlock situation happening similar to the SPARK-26265.
To fix this, we can move the `freePage` call in `loadNext` outside of `Synchronized` block similar to the fix in SPARK-26265
## How was this patch tested?
Manual tests were being done and will also try to add a test.
Closes#24265 from venkata91/deadlock-sorter.
Authored-by: Venkata krishnan Sowrirajan <vsowrirajan@qubole.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
HighlyCompressedMapStatus uses RoaringBitmap to record the empty blocks. But RoaringBitmap couldn't be ser/deser with unsafe KryoSerializer.
It's a bug of RoaringBitmap-0.5.11 and fixed in latest version.
This is an update of #24157
## How was this patch tested?
Add a UT
Closes#24264 from LantaoJin/SPARK-27216.
Lead-authored-by: LantaoJin <jinlantao@gmail.com>
Co-authored-by: Lantao Jin <jinlantao@gmail.com>
Signed-off-by: Imran Rashid <irashid@cloudera.com>
## What changes were proposed in this pull request?
Use Single Abstract Method syntax where possible (and minor related cleanup). Comments below. No logic should change here.
## How was this patch tested?
Existing tests.
Closes#24241 from srowen/SPARK-27323.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Update thread audit whitelist to skip threads of the global broadcast exchange thread pool, process reaper and Hadoop FS statistics data reference cleaner thread.
## How was this patch tested?
Via existing UT using broadcast exchange via `sbt` i.e:
```
> project sql
> testOnly *.SessionStateSuite -- -z "fork new sessions and run query on inherited table"
```
Before (wrapped long line for manually to save horizontal scrolling for reviewers):
```
===== POSSIBLE THREAD LEAK IN SUITE o.a.s.sql.SessionStateSuite,
thread names: broadcast-exchange-6, broadcast-exchange-0,
broadcast-exchange-2, broadcast-exchange-5, broadcast-exchange-7,
broadcast-exchange-4, broadcast-exchange-1, process reaper, broadcast-exchange-3,
org.apache.hadoop.fs.FileSystem$Statistics$StatisticsDataReferenceCleaner =====
```
After this change no possible thread leak detected.
Closes#24244 from attilapiros/thread-audit-minor.
Authored-by: “attilapiros” <piros.attila.zsolt@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
This PR is a FollowUp of https://github.com/apache/spark/pull/24196. It improves the test case by using the parameters that are being used in the actual scenarios.
## How was this patch tested?
N/A
Closes#24257 from gatorsmile/followupSPARK-27244.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Add AL2 license to metadata of all .md files.
This seemed to be the tidiest way as it will get ignored by .md renderers and other tools. Attempts to write them as markdown comments revealed that there is no such standard thing.
## How was this patch tested?
Doc build
Closes#24243 from srowen/SPARK-26918.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
spark.task.cpus should be less or equal than spark.executor.cores when use static executor allocation
## How was this patch tested?
manual
Closes#24131 from liutang123/SPARK-27192.
Authored-by: liulijia <liutang123@yeah.net>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
When logConf is set to true, config keys that contain password were printed in cleartext in driver log. This change uses the already present redact method in Utils, to redact all the passwords based on redact pattern in SparkConf and then print the conf to driver log thus ensuring that sensitive information like passwords is not printed in clear text.
## How was this patch tested?
This patch was tested through `SparkConfSuite` & then entire unit test through sbt
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#24196 from ninadingole/SPARK-27244.
Authored-by: Ninad Ingole <robert.wallis@example.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
I was playing with the scheduler and found this weird thing. In `TaskSchedulerImpl` we import `scala.collection.Set` without any reason. This is bad in practice, as it silently changes the actual class when we simply type `Set`, which by default should point to the immutable set.
This change only affects one method: `getExecutorsAliveOnHost`. I checked all the caller side and none of them need a general `Set` type.
## How was this patch tested?
N/A
Closes#24231 from cloud-fan/minor.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Raise the threshold size for serialized task size at which a warning is generated from 100KiB to 1000KiB.
As several people have noted, the original change for this JIRA highlighted that this threshold is low. Test output regularly shows:
```
- sorting on StringType with nullable=false, sortOrder=List('a DESC NULLS LAST)
22:47:53.320 WARN org.apache.spark.scheduler.TaskSetManager: Stage 80 contains a task of very large size (755 KiB). The maximum recommended task size is 100 KiB.
22:47:53.348 WARN org.apache.spark.scheduler.TaskSetManager: Stage 81 contains a task of very large size (755 KiB). The maximum recommended task size is 100 KiB.
22:47:53.417 WARN org.apache.spark.scheduler.TaskSetManager: Stage 83 contains a task of very large size (755 KiB). The maximum recommended task size is 100 KiB.
22:47:53.444 WARN org.apache.spark.scheduler.TaskSetManager: Stage 84 contains a task of very large size (755 KiB). The maximum recommended task size is 100 KiB.
...
- SPARK-20688: correctly check analysis for scalar sub-queries
22:49:10.314 WARN org.apache.spark.scheduler.DAGScheduler: Broadcasting large task binary with size 150.8 KiB
- SPARK-21835: Join in correlated subquery should be duplicateResolved: case 1
22:49:10.595 WARN org.apache.spark.scheduler.DAGScheduler: Broadcasting large task binary with size 150.7 KiB
22:49:10.744 WARN org.apache.spark.scheduler.DAGScheduler: Broadcasting large task binary with size 150.7 KiB
22:49:10.894 WARN org.apache.spark.scheduler.DAGScheduler: Broadcasting large task binary with size 150.7 KiB
- SPARK-21835: Join in correlated subquery should be duplicateResolved: case 2
- SPARK-21835: Join in correlated subquery should be duplicateResolved: case 3
- SPARK-23316: AnalysisException after max iteration reached for IN query
22:49:11.559 WARN org.apache.spark.scheduler.DAGScheduler: Broadcasting large task binary with size 154.2 KiB
```
It seems that a larger threshold of about 1MB is more suitable.
## How was this patch tested?
Existing tests.
Closes#24226 from srowen/SPARK-26660.2.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
## What changes were proposed in this pull request?
As per https://docs.oracle.com/javase/8/docs/api/java/lang/ClassLoader.html
``Class loaders that support concurrent loading of classes are known as parallel capable class loaders and are required to register themselves at their class initialization time by invoking the ClassLoader.registerAsParallelCapable method. Note that the ClassLoader class is registered as parallel capable by default. However, its subclasses still need to register themselves if they are parallel capable. ``
i.e we can have finer class loading locks by registering classloaders as parallel capable. (Refer to deadlock due to macro lock https://issues.apache.org/jira/browse/SPARK-26961).
All the classloaders we have are wrapper of URLClassLoader which by itself is parallel capable.
But this cannot be achieved by scala code due to static registration Refer https://github.com/scala/bug/issues/11429
## How was this patch tested?
All Existing UT must pass
Closes#24126 from ajithme/driverlock.
Authored-by: Ajith <ajith2489@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Currently, if we want to configure `spark.sql.files.maxPartitionBytes` to 256 megabytes, we must set `spark.sql.files.maxPartitionBytes=268435456`, which is very unfriendly to users.
And if we set it like this:`spark.sql.files.maxPartitionBytes=256M`, we will encounter this exception:
```
Exception in thread "main" java.lang.IllegalArgumentException:
spark.sql.files.maxPartitionBytes should be long, but was 256M
at org.apache.spark.internal.config.ConfigHelpers$.toNumber(ConfigBuilder.scala)
```
This PR use `bytesConf` to replace `longConf` or `intConf`, if the configuration is used to set the number of bytes.
Configuration change list:
`spark.files.maxPartitionBytes`
`spark.files.openCostInBytes`
`spark.shuffle.sort.initialBufferSize`
`spark.shuffle.spill.initialMemoryThreshold`
`spark.sql.autoBroadcastJoinThreshold`
`spark.sql.files.maxPartitionBytes`
`spark.sql.files.openCostInBytes`
`spark.sql.defaultSizeInBytes`
## How was this patch tested?
1.Existing unit tests
2.Manual testing
Closes#24187 from 10110346/bytesConf.
Authored-by: liuxian <liu.xian3@zte.com.cn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This applies some minor updates/cleaning following up SPARK-26928, notably renaming JVMCPU.scala to JVMCPUSource.scala.
## How was this patch tested?
Manually tested
Closes#24201 from LucaCanali/fixupSPARK-26928.
Authored-by: Luca Canali <luca.canali@cern.ch>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Remove Scala 2.11 support in build files and docs, and in various parts of code that accommodated 2.11. See some targeted comments below.
## How was this patch tested?
Existing tests.
Closes#23098 from srowen/SPARK-26132.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
The hashSeed method allocates 64 bytes instead of 8. Other bytes are always zeros (thanks to default behavior of ByteBuffer). And they could be excluded from hash calculation because they don't differentiate inputs.
## How was this patch tested?
By running the existing tests - XORShiftRandomSuite
Closes#20793 from MaxGekk/hash-buff-size.
Lead-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
For [SPARK-27184](https://issues.apache.org/jira/browse/SPARK-27184)
In the `org.apache.spark.internal.config`, we define the variables of `FILES` and `JARS`, we can use them instead of "spark.jars" and "spark.files".
```scala
private[spark] val JARS = ConfigBuilder("spark.jars")
.stringConf
.toSequence
.createWithDefault(Nil)
```
```scala
private[spark] val FILES = ConfigBuilder("spark.files")
.stringConf
.toSequence
.createWithDefault(Nil)
```
Other :
In the `org.apache.spark.internal.config`, we define the variables of `SUBMIT_PYTHON_FILES ` and `SUBMIT_DEPLOY_MODE `, we can use them instead of "spark.submit.pyFiles" and "spark.submit.deployMode".
```scala
private[spark] val SUBMIT_PYTHON_FILES = ConfigBuilder("spark.submit.pyFiles")
.stringConf
.toSequence
.createWithDefault(Nil)
```
```scala
private[spark] val SUBMIT_DEPLOY_MODE = ConfigBuilder("spark.submit.deployMode")
.stringConf
.createWithDefault("client")
```
Closes#24123 from hehuiyuan/hehuiyuan-patch-6.
Authored-by: hehuiyuan <hehuiyuan@ZBMAC-C02WD3K5H.local>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This patch fixes the issue that ClientEndpoint in standalone cluster doesn't recognize about driver options which are passed to SparkConf instead of system properties. When `Client` is executed via cli they should be provided as system properties, but with `spark-submit` they can be provided as SparkConf. (SpartSubmit will call `ClientApp.start` with SparkConf which would contain these options.)
## How was this patch tested?
Manually tested via following steps:
1) setup standalone cluster (launch master and worker via `./sbin/start-all.sh`)
2) submit one of example app with standalone cluster mode
```
./bin/spark-submit --class org.apache.spark.examples.SparkPi --master "spark://localhost:7077" --conf "spark.driver.extraJavaOptions=-Dfoo=BAR" --deploy-mode "cluster" --num-executors 1 --driver-memory 512m --executor-memory 512m --executor-cores 1 examples/jars/spark-examples*.jar 10
```
3) check whether `foo=BAR` is provided in system properties in Spark UI
<img width="877" alt="Screen Shot 2019-03-21 at 8 18 04 AM" src="https://user-images.githubusercontent.com/1317309/54728501-97db1700-4bc1-11e9-89da-078445c71e9b.png">
Closes#24163 from HeartSaVioR/SPARK-26606.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
This patch modifies Utils.classForName to have optional parameters - initialize, noSparkClassLoader - to let callers of Class.forName with thread context classloader to use it instead. This helps to reduce scalastyle off for Class.forName.
## How was this patch tested?
Existing UTs.
Closes#24148 from HeartSaVioR/MINOR-reduce-scalastyle-off-for-class-forname.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Refactored code in tests regarding the "withLogAppender()" pattern by creating a general helper method in SparkFunSuite.
## How was this patch tested?
Passed existing tests.
Closes#24172 from maryannxue/log-appender.
Authored-by: maryannxue <maryannxue@apache.org>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
[SPARK-26977](https://issues.apache.org/jira/browse/SPARK-26977) introduced very strange bug which spark-shell is no longer able to load classes which are provided via `--packages`. TBH I don't know about the details why it is broken, but looks like initializing `object class` brings the weirdness (maybe due to static initialization done twice?).
This patch removes the logic to leave warning log when main class is scala.App, to not deal with such complexity for just leaving warning message.
## How was this patch tested?
Manual test: suppose we run spark-shell with `--packages` option like below:
```
./bin/spark-shell --verbose --master "local[*]" --packages org.apache.spark:spark-sql-kafka-0-10_2.11:2.4.0
```
Before this patch, importing class in transitive dependency fails:
```
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
Spark context Web UI available at http://localhost:4040
Spark context available as 'sc' (master = local[*], app id = local-1553005771597).
Spark session available as 'spark'.
Welcome to
____ __
/ __/__ ___ _____/ /__
_\ \/ _ \/ _ `/ __/ '_/
/___/ .__/\_,_/_/ /_/\_\ version 3.0.0-SNAPSHOT
/_/
Using Scala version 2.12.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_191)
Type in expressions to have them evaluated.
Type :help for more information.
scala> import org.apache.kafka
<console>:23: error: object kafka is not a member of package org.apache
import org.apache.kafka
```
After this patch, importing class in transitive dependency succeeds:
```
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
Spark context Web UI available at http://localhost:4040
Spark context available as 'sc' (master = local[*], app id = local-1553004095542).
Spark session available as 'spark'.
Welcome to
____ __
/ __/__ ___ _____/ /__
_\ \/ _ \/ _ `/ __/ '_/
/___/ .__/\_,_/_/ /_/\_\ version 3.0.0-SNAPSHOT
/_/
Using Scala version 2.12.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_191)
Type in expressions to have them evaluated.
Type :help for more information.
scala> import org.apache.kafka
import org.apache.kafka
```
Closes#24147 from HeartSaVioR/SPARK-27205.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
```scala
val KRYO_USE_UNSAFE = ConfigBuilder("spark.kyro.unsafe")
.booleanConf
.createWithDefault(false)
val KRYO_USE_POOL = ConfigBuilder("spark.kyro.pool")
.booleanConf
.createWithDefault(true)
```
**kyro should be kryo**
## How was this patch tested?
no need
Closes#24156 from LantaoJin/SPARK-27215.
Authored-by: Lantao Jin <jinlantao@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
Environment Page in SparkUI have all the configuration sorted by key. But this is not the case in History server case, to keep UX same, we can have it sorted in history server too
## What changes were proposed in this pull request?
On render of Env page the properties are sorted before creating page
## How was this patch tested?
Manually tested in UI
Closes#24143 from ajithme/historyenv.
Authored-by: Ajith <ajith2489@gmail.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
As we all know that spark on Yarn uses DB https://github.com/apache/spark/pull/7943 to record RegisteredExecutors information which can be reloaded and used again when the ExternalShuffleService is restarted .
The RegisteredExecutors information can't be recorded both in the mode of spark's standalone and spark on k8s , which will cause the RegisteredExecutors information to be lost ,when the ExternalShuffleService is restarted.
To solve the problem above, a method is proposed and is committed .
## How was this patch tested?
new unit tests
Closes#23393 from weixiuli/SPARK-26288.
Authored-by: weixiuli <weixiuli@jd.com>
Signed-off-by: Imran Rashid <irashid@cloudera.com>
…when trying to kill executors either due to dynamic allocation or blacklisting
## What changes were proposed in this pull request?
There are two deadlocks as a result of the interplay between three different threads:
**task-result-getter thread**
**spark-dynamic-executor-allocation thread**
**dispatcher-event-loop thread(makeOffers())**
The fix ensures ordering synchronization constraint by acquiring lock on `TaskSchedulerImpl` before acquiring lock on `CoarseGrainedSchedulerBackend` in `makeOffers()` as well as killExecutors() method. This ensures resource ordering between the threads and thus, fixes the deadlocks.
## How was this patch tested?
Manual Tests
Closes#24072 from pgandhi999/SPARK-27112-2.
Authored-by: pgandhi <pgandhi@verizonmedia.com>
Signed-off-by: Imran Rashid <irashid@cloudera.com>
## What changes were proposed in this pull request?
When Result stage has zero tasks, the Job End event is never fired, hence the Job is always running in UI. Example: sc.emptyRDD[Int].countApprox(1000) never finishes even it has no tasks to launch
## How was this patch tested?
Added UT
Closes#24100 from ajithme/emptyRDD.
Authored-by: Ajith <ajith2489@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
This time tested against Scala 2.11 as well
Closes#24116 from fitermay/master.
Authored-by: fitermay <fiterman@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
When we run YarnSchedulerBackendSuite, the class path seems to be made from the classes folder(resource-managers/yarn/target/scala-2.12/classes) instead of jar (resource-managers/yarn/target/spark-yarn_2.12-3.0.0-SNAPSHOT.jar) . ui.getHandlers is in spark-core and its loaded from spark-core.jar which is shaded and hence refers to org.spark_project.jetty.servlet.ServletContextHandler
Here in org.apache.spark.scheduler.cluster.YarnSchedulerBackend, as its not shaded, it expects org.eclipse.jetty.servlet.ServletContextHandler
Refer discussion https://issues.apache.org/jira/browse/SPARK-27122?focusedCommentId=16792318&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16792318
Hence as a fix, org.apache.spark.ui.WebUI must only return a wrapper class instance or references so that Jetty classes can be avoided in getters which are accessed outside spark-core
## How was this patch tested?
Existing UT can pass
Closes#24088 from ajithme/shadebug.
Authored-by: Ajith <ajith2489@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
this pr add check this spark.diskStore.subDirectories > 0.This value need to be checked before it can be used.
## How was this patch tested?
N/A
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#24024 from lcqzte10192193/wid-lcq-190308.
Authored-by: lichaoqun <li.chaoqun@zte.com.cn>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Currently, when enabled streaming dynamic allocation for streaming applications, the maxNumExecutorFailures in ApplicationMaster is still computed with `spark.dynamicAllocation.maxExecutors`.
Actually, we should consider `spark.streaming.dynamicAllocation.maxExecutors` instead.
Related codes:
f87153a3ac/resource-managers/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala (L101)
## How was this patch tested?
NA
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#23845 from liupc/Fix-incorrect-maxNumExecutorFailures-for-streaming.
Lead-authored-by: Liupengcheng <liupengcheng@xiaomi.com>
Co-authored-by: liupengcheng <liupengcheng@xiaomi.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Executor heartbeat interval is configurable by `"spark.executor.heartbeatInterval"`. But in a comment, heartbeat interval is presented as a constant `10s`. This pr tries to correct the description.
## How was this patch tested?
Existing unit tests.
Closes#24101 from SongYadong/heartbeat_interval_comment.
Authored-by: SongYadong <song.yadong1@zte.com.cn>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
