## What changes were proposed in this pull request?
Jira : https://issues.apache.org/jira/browse/SPARK-17451
`CoarseGrainedExecutorBackend` in some failure cases exits the JVM. While this does not have any issue, from the driver UI there is no specific reason captured for this. In this PR, I am adding functionality to `exitExecutor` to notify driver that the executor is exiting.
## How was this patch tested?
Ran the change over a test env and took down shuffle service before the executor could register to it. In the driver logs, where the job failure reason is mentioned (ie. `Job aborted due to stage ...` it gives the correct reason:
Before:
`ExecutorLostFailure (executor ZZZZZZZZZ exited caused by one of the running tasks) Reason: Remote RPC client disassociated. Likely due to containers exceeding thresholds, or network issues. Check driver logs for WARN messages.`
After:
`ExecutorLostFailure (executor ZZZZZZZZZ exited caused by one of the running tasks) Reason: Unable to create executor due to java.util.concurrent.TimeoutException: Timeout waiting for task.`
Author: Tejas Patil <tejasp@fb.com>
Closes#15013 from tejasapatil/SPARK-17451_inform_driver.
## What changes were proposed in this pull request?
Following https://github.com/apache/spark/pull/14969 for some reason the MiMa excludes weren't complete, but still passed the PR builder. This adds 3 more excludes from https://amplab.cs.berkeley.edu/jenkins/view/Spark%20QA%20Test%20(Dashboard)/job/spark-master-test-sbt-hadoop-2.2/1749/consoleFull
It also moves the excludes to their own Seq in the build, as they probably should have been.
Even though this is merged to 2.1.x only / master, I left the exclude in for 2.0.x in case we back port. It's a private API so is always a false positive.
## How was this patch tested?
Jenkins build
Author: Sean Owen <sowen@cloudera.com>
Closes#15110 from srowen/SPARK-17406.2.
## What changes were proposed in this pull request?
Optimize a while loop during batch inserts
## How was this patch tested?
Unit tests were done, specifically "mvn test" for sql
Author: John Muller <jmuller@us.imshealth.com>
Closes#15098 from blue666man/SPARK-17536.
## What changes were proposed in this pull request?
The job page will be too slow to open when there are thousands of executor events(added or removed). I found that in ExecutorsTab file, executorIdToData will not remove elements, it will increase all the time.Before this pr, it looks like [timeline1.png](https://issues.apache.org/jira/secure/attachment/12827112/timeline1.png). After this pr, it looks like [timeline2.png](https://issues.apache.org/jira/secure/attachment/12827113/timeline2.png)(we can set how many executor events will be displayed)
Author: cenyuhai <cenyuhai@didichuxing.com>
Closes#14969 from cenyuhai/SPARK-17406.
## What changes were proposed in this pull request?
when i use sc.makeRDD below
```
val data3 = sc.makeRDD(Seq())
println(data3.partitions.length)
```
I got an error:
Exception in thread "main" java.lang.IllegalArgumentException: Positive number of slices required
We can fix this bug just modify the last line ,do a check of seq.size
```
def makeRDD[T: ClassTag](seq: Seq[(T, Seq[String])]): RDD[T] = withScope {
assertNotStopped()
val indexToPrefs = seq.zipWithIndex.map(t => (t._2, t._1._2)).toMap
new ParallelCollectionRDD[T](this, seq.map(_._1), math.max(seq.size, defaultParallelism), indexToPrefs)
}
```
## How was this patch tested?
manual tests
(If this patch involves UI changes, please attach a screenshot; otherwise, remove this)
Author: codlife <1004910847@qq.com>
Author: codlife <wangjianfei15@otcaix.iscas.ac.cn>
Closes#15077 from codlife/master.
## What changes were proposed in this pull request?
This PR has the appendRowUntilExceedingPageSize test in RowBasedKeyValueBatchSuite use whatever spark.buffer.pageSize value a user has specified to prevent a test failure for anyone testing Apache Spark on a box with a reduced page size. The test is currently hardcoded to use the default page size which is 64 MB so this minor PR is a test improvement
## How was this patch tested?
Existing unit tests with 1 MB page size and with 64 MB (the default) page size
Author: Adam Roberts <aroberts@uk.ibm.com>
Closes#15079 from a-roberts/patch-5.
## What changes were proposed in this pull request?
as the TODO described,
check weight vector size and if wrong throw exception.
## How was this patch tested?
existing tests.
Author: WeichenXu <WeichenXu123@outlook.com>
Closes#15060 from WeichenXu123/check_input_weight_size_of_ann.
### What changes were proposed in this pull request?
For the following `ALTER TABLE` DDL, we should issue an exception when the target table is a `VIEW`:
```SQL
ALTER TABLE viewName SET LOCATION '/path/to/your/lovely/heart'
ALTER TABLE viewName SET SERDE 'whatever'
ALTER TABLE viewName SET SERDEPROPERTIES ('x' = 'y')
ALTER TABLE viewName PARTITION (a=1, b=2) SET SERDEPROPERTIES ('x' = 'y')
ALTER TABLE viewName ADD IF NOT EXISTS PARTITION (a='4', b='8')
ALTER TABLE viewName DROP IF EXISTS PARTITION (a='2')
ALTER TABLE viewName RECOVER PARTITIONS
ALTER TABLE viewName PARTITION (a='1', b='q') RENAME TO PARTITION (a='100', b='p')
```
In addition, `ALTER TABLE RENAME PARTITION` is unable to handle data source tables, just like the other `ALTER PARTITION` commands. We should issue an exception instead.
### How was this patch tested?
Added a few test cases.
Author: gatorsmile <gatorsmile@gmail.com>
Closes#15004 from gatorsmile/altertable.
The expression like `if (memoryMap(taskAttemptId) == 0) memoryMap.remove(taskAttemptId)` in method `releaseUnrollMemoryForThisTask` and `releasePendingUnrollMemoryForThisTask` should be called after release memory operation, whatever `memoryToRelease` is > 0 or not.
If the memory of a task has been set to 0 when calling a `releaseUnrollMemoryForThisTask` or a `releasePendingUnrollMemoryForThisTask` method, the key in the memory map corresponding to that task will never be removed from the hash map.
See the details in [SPARK-17465](https://issues.apache.org/jira/browse/SPARK-17465).
Author: Xing SHI <shi-kou@indetail.co.jp>
Closes#15022 from saturday-shi/SPARK-17465.
## What changes were proposed in this pull request?
For large objects, pickle does not raise useful error messages. However, we can wrap them to be slightly more user friendly:
Example 1:
```
def run():
import numpy.random as nr
b = nr.bytes(8 * 1000000000)
sc.parallelize(range(1000), 1000).map(lambda x: len(b)).count()
run()
```
Before:
```
error: 'i' format requires -2147483648 <= number <= 2147483647
```
After:
```
pickle.PicklingError: Object too large to serialize: 'i' format requires -2147483648 <= number <= 2147483647
```
Example 2:
```
def run():
import numpy.random as nr
b = sc.broadcast(nr.bytes(8 * 1000000000))
sc.parallelize(range(1000), 1000).map(lambda x: len(b.value)).count()
run()
```
Before:
```
SystemError: error return without exception set
```
After:
```
cPickle.PicklingError: Could not serialize broadcast: SystemError: error return without exception set
```
## How was this patch tested?
Manually tried out these cases
cc davies
Author: Eric Liang <ekl@databricks.com>
Closes#15026 from ericl/spark-17472.
## What changes were proposed in this pull request?
Make CollectionAccumulator and SetAccumulator's value can be read thread-safely to fix the ConcurrentModificationException reported in [JIRA](https://issues.apache.org/jira/browse/SPARK-17463).
## How was this patch tested?
Existing tests.
Author: Shixiong Zhu <shixiong@databricks.com>
Closes#15063 from zsxwing/SPARK-17463.
## What changes were proposed in this pull request?
Due to race conditions, the ` assert(numExecutorsRunning <= targetNumExecutors)` can fail causing `AssertionError`. So removed the assertion, instead moved the conditional check before launching new container:
```
java.lang.AssertionError: assertion failed
at scala.Predef$.assert(Predef.scala:156)
at org.apache.spark.deploy.yarn.YarnAllocator$$anonfun$runAllocatedContainers$1.org$apache$spark$deploy$yarn$YarnAllocator$$anonfun$$updateInternalState$1(YarnAllocator.scala:489)
at org.apache.spark.deploy.yarn.YarnAllocator$$anonfun$runAllocatedContainers$1$$anon$1.run(YarnAllocator.scala:519)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
```
## How was this patch tested?
This was manually tested using a large ForkAndJoin job with Dynamic Allocation enabled to validate the failing job succeeds, without any such exception.
Author: Kishor Patil <kpatil@yahoo-inc.com>
Closes#15069 from kishorvpatil/SPARK-17511.
## What changes were proposed in this pull request?
Currently, ORDER BY clause returns nulls value according to sorting order (ASC|DESC), considering null value is always smaller than non-null values.
However, SQL2003 standard support NULLS FIRST or NULLS LAST to allow users to specify whether null values should be returned first or last, regardless of sorting order (ASC|DESC).
This PR is to support this new feature.
## How was this patch tested?
New test cases are added to test NULLS FIRST|LAST for regular select queries and windowing queries.
(If this patch involves UI changes, please attach a screenshot; otherwise, remove this)
Author: Xin Wu <xinwu@us.ibm.com>
Closes#14842 from xwu0226/SPARK-10747.
## What changes were proposed in this pull request?
I first thought they are missing because they are kind of hidden options but it seems they are just missing.
For example, `spark.sql.parquet.mergeSchema` is documented in [sql-programming-guide.md](https://github.com/apache/spark/blob/master/docs/sql-programming-guide.md) but this function is missing whereas many options such as `spark.sql.join.preferSortMergeJoin` are not documented but have its own function individually.
So, this PR suggests making them consistent by adding the missing functions for some options in `SQLConf` and use them where applicable, in order to make them more readable.
## How was this patch tested?
Existing tests should cover this.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#14678 from HyukjinKwon/sqlconf-cleanup.
## What changes were proposed in this pull request?
In PySpark, `df.take(1)` runs a single-stage job which computes only one partition of the DataFrame, while `df.limit(1).collect()` computes all partitions and runs a two-stage job. This difference in performance is confusing.
The reason why `limit(1).collect()` is so much slower is that `collect()` internally maps to `df.rdd.<some-pyspark-conversions>.toLocalIterator`, which causes Spark SQL to build a query where a global limit appears in the middle of the plan; this, in turn, ends up being executed inefficiently because limits in the middle of plans are now implemented by repartitioning to a single task rather than by running a `take()` job on the driver (this was done in #7334, a patch which was a prerequisite to allowing partition-local limits to be pushed beneath unions, etc.).
In order to fix this performance problem I think that we should generalize the fix from SPARK-10731 / #8876 so that `DataFrame.collect()` also delegates to the Scala implementation and shares the same performance properties. This patch modifies `DataFrame.collect()` to first collect all results to the driver and then pass them to Python, allowing this query to be planned using Spark's `CollectLimit` optimizations.
## How was this patch tested?
Added a regression test in `sql/tests.py` which asserts that the expected number of jobs, stages, and tasks are run for both queries.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#15068 from JoshRosen/pyspark-collect-limit.
### What changes were proposed in this pull request?
As explained in https://github.com/apache/spark/pull/14797:
>Some analyzer rules have assumptions on logical plans, optimizer may break these assumption, we should not pass an optimized query plan into QueryExecution (will be analyzed again), otherwise we may some weird bugs.
For example, we have a rule for decimal calculation to promote the precision before binary operations, use PromotePrecision as placeholder to indicate that this rule should not apply twice. But a Optimizer rule will remove this placeholder, that break the assumption, then the rule applied twice, cause wrong result.
We should not optimize the query in CTAS more than once. For example,
```Scala
spark.range(99, 101).createOrReplaceTempView("tab1")
val sqlStmt = "SELECT id, cast(id as long) * cast('1.0' as decimal(38, 18)) as num FROM tab1"
sql(s"CREATE TABLE tab2 USING PARQUET AS $sqlStmt")
checkAnswer(spark.table("tab2"), sql(sqlStmt))
```
Before this PR, the results do not match
```
== Results ==
!== Correct Answer - 2 == == Spark Answer - 2 ==
![100,100.000000000000000000] [100,null]
[99,99.000000000000000000] [99,99.000000000000000000]
```
After this PR, the results match.
```
+---+----------------------+
|id |num |
+---+----------------------+
|99 |99.000000000000000000 |
|100|100.000000000000000000|
+---+----------------------+
```
In this PR, we do not treat the `query` in CTAS as a child. Thus, the `query` will not be optimized when optimizing CTAS statement. However, we still need to analyze it for normalizing and verifying the CTAS in the Analyzer. Thus, we do it in the analyzer rule `PreprocessDDL`, because so far only this rule needs the analyzed plan of the `query`.
### How was this patch tested?
Added a test
Author: gatorsmile <gatorsmile@gmail.com>
Closes#15048 from gatorsmile/ctasOptimized.
## What changes were proposed in this pull request?
Point references to spark-packages.org to https://cwiki.apache.org/confluence/display/SPARK/Third+Party+Projects
This will be accompanied by a parallel change to the spark-website repo, and additional changes to this wiki.
## How was this patch tested?
Jenkins tests.
Author: Sean Owen <sowen@cloudera.com>
Closes#15075 from srowen/SPARK-17445.
## What changes were proposed in this pull request?
Scala's List.length method is O(N) and it makes the gatherCompressibilityStats function O(N^2). Eliminate the List.length calls by writing it in Scala way.
https://github.com/scala/scala/blob/2.10.x/src/library/scala/collection/LinearSeqOptimized.scala#L36
As suggested. Extended the fix to HiveInspectors and AggregationIterator classes as well.
## How was this patch tested?
Profiled a Spark job and found that CompressibleColumnBuilder is using 39% of the CPU. Out of this 39% CompressibleColumnBuilder->gatherCompressibilityStats is using 23% of it. 6.24% of the CPU is spend on List.length which is called inside gatherCompressibilityStats.
After this change we started to save 6.24% of the CPU.
Author: Ergin Seyfe <eseyfe@fb.com>
Closes#15032 from seyfe/gatherCompressibilityStats.
## What changes were proposed in this pull request?
In the comment, there is redundant `the estimated`.
This PR simply remove the redundant comment and adjusts format.
Author: wm624@hotmail.com <wm624@hotmail.com>
Closes#15091 from wangmiao1981/comment.
## What changes were proposed in this pull request?
This pull request removes the SparkContext.clearFiles() method from the PySpark API as the method was removed from the Scala API in 8ce645d4ee. Using that method in PySpark leads to an exception as PySpark tries to call the non-existent method on the JVM side.
## How was this patch tested?
Existing tests (though none of them tested this particular method).
Author: Sami Jaktholm <sjakthol@outlook.com>
Closes#15081 from sjakthol/pyspark-sc-clearfiles.
## What changes were proposed in this pull request?
The relation between spark.network.timeout and spark.executor.heartbeatInterval should be mentioned in the document.
… network timeout]
Author: Jagadeesan <as2@us.ibm.com>
Closes#15042 from jagadeesanas2/SPARK-17449.
## What changes were proposed in this pull request?
This PR tries to add a SparkR vignette, which works as a friendly guidance going through the functionality provided by SparkR.
## How was this patch tested?
Manual test.
Author: junyangq <qianjunyang@gmail.com>
Author: Shivaram Venkataraman <shivaram@cs.berkeley.edu>
Author: Junyang Qian <junyangq@databricks.com>
Closes#14980 from junyangq/SPARKR-vignette.
## What changes were proposed in this pull request?
If a user provides listeners inside the Hive Conf, the configuration for these listeners are passed to the Hive Execution Client as well. This may cause issues for two reasons:
1. The Execution Client will actually generate garbage
2. The listener class needs to be both in the Spark Classpath and Hive Classpath
This PR empties the listener configurations in `HiveUtils.newTemporaryConfiguration` so that the execution client will not contain the listener confs, but the metadata client will.
## How was this patch tested?
Unit tests
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#15086 from brkyvz/null-listeners.
## What changes were proposed in this pull request?
In `ReorderAssociativeOperator` rule, we extract foldable expressions with Add/Multiply arithmetics, and replace with eval literal. For example, `(a + 1) + (b + 2)` is optimized to `(a + b + 3)` by this rule.
For aggregate operator, output expressions should be derived from groupingExpressions, current implemenation of `ReorderAssociativeOperator` rule may break this promise. A instance could be:
```
SELECT
((t1.a + 1) + (t2.a + 2)) AS out_col
FROM
testdata2 AS t1
INNER JOIN
testdata2 AS t2
ON
(t1.a = t2.a)
GROUP BY (t1.a + 1), (t2.a + 2)
```
`((t1.a + 1) + (t2.a + 2))` is optimized to `(t1.a + t2.a + 3)`, which could not be derived from `ExpressionSet((t1.a +1), (t2.a + 2))`.
Maybe we should improve the rule of `ReorderAssociativeOperator` by adding a GroupingExpressionSet to keep Aggregate.groupingExpressions, and respect these expressions during the optimize stage.
## How was this patch tested?
Add new test case in `ReorderAssociativeOperatorSuite`.
Author: jiangxingbo <jiangxb1987@gmail.com>
Closes#14917 from jiangxb1987/rao.
## What changes were proposed in this pull request?
CollectLimit.execute() incorrectly omits per-partition limits, leading to performance regressions in case this case is hit (which should not happen in normal operation, but can occur in some cases (see #15068 for one example).
## How was this patch tested?
Regression test in SQLQuerySuite that asserts the number of records scanned from the input RDD.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#15070 from JoshRosen/SPARK-17515.
## What changes were proposed in this pull request?
This PR proposes to close some stale PRs and ones suggested to be closed by committer(s)
Closes#10052Closes#11079Closes#12661Closes#12772Closes#12958Closes#12990Closes#13409Closes#13779Closes#13811Closes#14577Closes#14714Closes#14875Closes#15020
## How was this patch tested?
N/A
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#15057 from HyukjinKwon/closing-stale-pr.
## What changes were proposed in this pull request?
When there is any Python UDF in the Project between Sort and Limit, it will be collected into TakeOrderedAndProjectExec, ExtractPythonUDFs failed to pull the Python UDFs out because QueryPlan.expressions does not include the expression inside Option[Seq[Expression]].
Ideally, we should fix the `QueryPlan.expressions`, but tried with no luck (it always run into infinite loop). In PR, I changed the TakeOrderedAndProjectExec to no use Option[Seq[Expression]] to workaround it. cc JoshRosen
## How was this patch tested?
Added regression test.
Author: Davies Liu <davies@databricks.com>
Closes#15030 from davies/all_expr.
## What changes were proposed in this pull request?
In Spark's `RDD.getOrCompute` we first try to read a local copy of a cached RDD block, then a remote copy, and only fall back to recomputing the block if no cached copy (local or remote) can be read. This logic works correctly in the case where no remote copies of the block exist, but if there _are_ remote copies and reads of those copies fail (due to network issues or internal Spark bugs) then the BlockManager will throw a `BlockFetchException` that will fail the task (and which could possibly fail the whole job if the read failures keep occurring).
In the cases of TorrentBroadcast and task result fetching we really do want to fail the entire job in case no remote blocks can be fetched, but this logic is inappropriate for reads of cached RDD blocks because those can/should be recomputed in case cached blocks are unavailable.
Therefore, I think that the `BlockManager.getRemoteBytes()` method should never throw on remote fetch errors and, instead, should handle failures by returning `None`.
## How was this patch tested?
Block manager changes should be covered by modified tests in `BlockManagerSuite`: the old tests expected exceptions to be thrown on failed remote reads, while the modified tests now expect `None` to be returned from the `getRemote*` method.
I also manually inspected all usages of `BlockManager.getRemoteValues()`, `getRemoteBytes()`, and `get()` to verify that they correctly pattern-match on the result and handle `None`. Note that these `None` branches are already exercised because the old `getRemoteBytes` returned `None` when no remote locations for the block could be found (which could occur if an executor died and its block manager de-registered with the master).
Author: Josh Rosen <joshrosen@databricks.com>
Closes#15037 from JoshRosen/SPARK-17485.
This patch makes a handful of post-Spark-2.0 MiMa exclusion and build updates. It should be merged to master and a subset of it should be picked into branch-2.0 in order to test Spark 2.0.1-SNAPSHOT.
- Remove the ` sketch`, `mllibLocal`, and `streamingKafka010` from the list of excluded subprojects so that MiMa checks them.
- Remove now-unnecessary special-case handling of the Kafka 0.8 artifact in `mimaSettings`.
- Move the exclusion added in SPARK-14743 from `v20excludes` to `v21excludes`, since that patch was only merged into master and not branch-2.0.
- Add exclusions for an API change introduced by SPARK-17096 / #14675.
- Add missing exclusions for the `o.a.spark.internal` and `o.a.spark.sql.internal` packages.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#15061 from JoshRosen/post-2.0-mima-changes.
This patch makes three minor refactorings to the BlockManager:
- Move the `if (info.tellMaster)` check out of `reportBlockStatus`; this fixes an issue where a debug logging message would incorrectly claim to have reported a block status to the master even though no message had been sent (in case `info.tellMaster == false`). This also makes it easier to write code which unconditionally sends block statuses to the master (which is necessary in another patch of mine).
- Split `removeBlock()` into two methods, the existing method and an internal `removeBlockInternal()` method which is designed to be called by internal code that already holds a write lock on the block. This is also needed by a followup patch.
- Instead of calling `getCurrentBlockStatus()` in `removeBlock()`, just pass `BlockStatus.empty`; the block status should always be empty following complete removal of a block.
These changes were originally authored as part of a bug fix patch which is targeted at branch-2.0 and master; I've split them out here into their own separate PR in order to make them easier to review and so that the behavior-changing parts of my other patch can be isolated to their own PR.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#15036 from JoshRosen/cache-failure-race-conditions-refactorings-only.
## What changes were proposed in this pull request?
MemoryStore may throw OutOfMemoryError when trying to cache a super big RDD that cannot fit in memory.
```
scala> sc.parallelize(1 to 1000000000, 100).map(x => new Array[Long](1000)).cache().count()
java.lang.OutOfMemoryError: Java heap space
at $line14.$read$$iw$$iw$$iw$$iw$$iw$$iw$$iw$$iw$$anonfun$1.apply(<console>:24)
at $line14.$read$$iw$$iw$$iw$$iw$$iw$$iw$$iw$$iw$$anonfun$1.apply(<console>:23)
at scala.collection.Iterator$$anon$11.next(Iterator.scala:409)
at scala.collection.Iterator$JoinIterator.next(Iterator.scala:232)
at org.apache.spark.storage.memory.PartiallyUnrolledIterator.next(MemoryStore.scala:683)
at org.apache.spark.InterruptibleIterator.next(InterruptibleIterator.scala:43)
at org.apache.spark.util.Utils$.getIteratorSize(Utils.scala:1684)
at org.apache.spark.rdd.RDD$$anonfun$count$1.apply(RDD.scala:1134)
at org.apache.spark.rdd.RDD$$anonfun$count$1.apply(RDD.scala:1134)
at org.apache.spark.SparkContext$$anonfun$runJob$5.apply(SparkContext.scala:1915)
at org.apache.spark.SparkContext$$anonfun$runJob$5.apply(SparkContext.scala:1915)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:70)
at org.apache.spark.scheduler.Task.run(Task.scala:86)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:274)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
```
Spark MemoryStore uses SizeTrackingVector as a temporary unrolling buffer to store all input values that it has read so far before transferring the values to storage memory cache. The problem is that when the input RDD is too big for caching in memory, the temporary unrolling memory SizeTrackingVector is not garbage collected in time. As SizeTrackingVector can occupy all available storage memory, it may cause the executor JVM to run out of memory quickly.
More info can be found at https://issues.apache.org/jira/browse/SPARK-17503
## How was this patch tested?
Unit test and manual test.
### Before change
Heap memory consumption
<img width="702" alt="screen shot 2016-09-12 at 4 16 15 pm" src="https://cloud.githubusercontent.com/assets/2595532/18429524/60d73a26-7906-11e6-9768-6f286f5c58c8.png">
Heap dump
<img width="1402" alt="screen shot 2016-09-12 at 4 34 19 pm" src="https://cloud.githubusercontent.com/assets/2595532/18429577/cbc1ef20-7906-11e6-847b-b5903f450b3b.png">
### After change
Heap memory consumption
<img width="706" alt="screen shot 2016-09-12 at 4 29 10 pm" src="https://cloud.githubusercontent.com/assets/2595532/18429503/4abe9342-7906-11e6-844a-b2f815072624.png">
Author: Sean Zhong <seanzhong@databricks.com>
Closes#15056 from clockfly/memory_store_leak.
## What changes were proposed in this pull request?
In order to avoid confusing user,
error message in `PairRDDfunctions`
`Default partitioner cannot partition array keys.`
is updated,
the one in `partitionBy` is replaced with
`Specified partitioner cannot partition array keys.`
other is replaced with
`Specified or default partitioner cannot partition array keys.`
## How was this patch tested?
N/A
Author: WeichenXu <WeichenXu123@outlook.com>
Closes#15045 from WeichenXu123/fix_partitionBy_error_message.
Code is equivalent, but map comprehency is most of the time faster than a map.
Author: Gaetan Semet <gaetan@xeberon.net>
Closes#14863 from Stibbons/map_comprehension.
## What changes were proposed in this pull request?
if there are many rdds in some situations,the sort will loss he performance servely,actually we needn't sort the rdds , we can just scan the rdds one time to gain the same goal.
## How was this patch tested?
manual tests
Author: codlife <1004910847@qq.com>
Closes#15039 from codlife/master.
The `TaskMetricsUIData.updatedBlockStatuses` field is assigned to but never read, increasing the memory consumption of the web UI. We should remove this field.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#15038 from JoshRosen/remove-updated-block-statuses-from-TaskMetricsUIData.
## What changes were proposed in this pull request?
This is a trivial patch that catches all `OutOfMemoryError` while building the broadcast hash relation and rethrows it by wrapping it in a nice error message.
## How was this patch tested?
Existing Tests
Author: Sameer Agarwal <sameerag@cs.berkeley.edu>
Closes#14979 from sameeragarwal/broadcast-join-error.
## What changes were proposed in this pull request?
#14956 reduced default k-means|| init steps to 2 from 5 only for spark.mllib package, we should also do same change for spark.ml and PySpark.
## How was this patch tested?
Existing tests.
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#15050 from yanboliang/spark-17389.
## What changes were proposed in this pull request?
During startup of Spark standalone, the script file spark-config.sh appends to the PYTHONPATH and can be sourced many times, causing duplicates in the path. This change adds a env flag that is set when the PYTHONPATH is appended so it will happen only one time.
## How was this patch tested?
Manually started standalone master/worker and verified PYTHONPATH has no duplicate entries.
Author: Bryan Cutler <cutlerb@gmail.com>
Closes#15028 from BryanCutler/fix-duplicate-pythonpath-SPARK-17336.
## What changes were proposed in this pull request?
Check the database warehouse used in Spark UT, and remove the existing database file before run the UT (SPARK-8368).
## How was this patch tested?
Run Spark UT with the command for several times:
./build/sbt -Pyarn -Phadoop-2.6 -Phive -Phive-thriftserver "test-only *HiveSparkSubmitSuit*"
Without the patch, the test case can be passed only at the first time, and always failed from the second time.
With the patch the test case always can be passed correctly.
Author: tone-zhang <tone.zhang@linaro.org>
Closes#14894 from tone-zhang/issue1.
## What changes were proposed in this pull request?
This PR build on #14976 and fixes a correctness bug that would cause the wrong quantile to be returned for small target errors.
## How was this patch tested?
This PR adds 8 unit tests that were failing without the fix.
Author: Timothy Hunter <timhunter@databricks.com>
Author: Sean Owen <sowen@cloudera.com>
Closes#15002 from thunterdb/ml-1783.
## What changes were proposed in this pull request?
Reduce default k-means|| init steps to 2 from 5. See JIRA for discussion.
See also https://github.com/apache/spark/pull/14948
## How was this patch tested?
Existing tests.
Author: Sean Owen <sowen@cloudera.com>
Closes#14956 from srowen/SPARK-17389.2.
## What changes were proposed in this pull request?
Fix summary() method's `return` description for spark.mlp
## How was this patch tested?
Ran tests locally on my laptop.
Author: Xin Ren <iamshrek@126.com>
Closes#15015 from keypointt/SPARK-16445-2.
## What changes were proposed in this pull request?
Share the ForkJoinTaskSupport between UnionRDD instances to avoid creating a huge number of threads if lots of RDDs are created at the same time.
## How was this patch tested?
This uses existing UnionRDD tests.
Author: Ryan Blue <blue@apache.org>
Closes#14985 from rdblue/SPARK-17396-use-shared-pool.
## What changes were proposed in this pull request?
#13584 resolved the issue of features and label columns conflict with ```RFormula``` default ones when loading libsvm data, but it still left some issues should be resolved:
1, It’s not necessary to check and rename label column.
Since we have considerations on the design of ```RFormula```, it can handle the case of label column already exists(with restriction of the existing label column should be numeric/boolean type). So it’s not necessary to change the column name to avoid conflict. If the label column is not numeric/boolean type, ```RFormula``` will throw exception.
2, We should rename features column name to new one if there is conflict, but appending a random value is enough since it was used internally only. We done similar work when implementing ```SQLTransformer```.
3, We should set correct new features column for the estimators. Take ```GLM``` as example:
```GLM``` estimator should set features column with the changed one(rFormula.getFeaturesCol) rather than the default “features”. Although it’s same when training model, but it involves problems when predicting. The following is the prediction result of GLM before this PR:
We should drop the internal used feature column name, otherwise, it will appear on the prediction DataFrame which will confused users. And this behavior is same as other scenarios which does not exist column name conflict.
After this PR:
## How was this patch tested?
Existing unit tests.
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#14993 from yanboliang/spark-15509.
(Updated version of [PR-9457](https://github.com/apache/spark/pull/9457), rebased on latest Spark master, and using mllib-local).
This implements a parallel version of personalized pagerank, which runs all propagations for a list of source vertices in parallel.
I ran a few benchmarks on the full [DBpedia](http://dbpedia.org/) graph. When running personalized pagerank for only one source node, the existing implementation is twice as fast as the parallel one (because of the SparseVector overhead). However for 10 source nodes, the parallel implementation is four times as fast. When increasing the number of source nodes, this difference becomes even greater.
Author: Yves Raimond <yraimond@netflix.com>
Closes#14998 from moustaki/parallel-ppr.
## What changes were proposed in this pull request?
This PR fixes `ColumnVectorUtils.populate` so that Parquet vectorized reader can read partitioned table with dates/timestamps. This works fine with Parquet normal reader.
This is being only called within [VectorizedParquetRecordReader.java#L185](https://github.com/apache/spark/blob/master/sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedParquetRecordReader.java#L185).
When partition column types are explicitly given to `DateType` or `TimestampType` (rather than inferring the type of partition column), this fails with the exception below:
```
16/09/01 10:30:07 ERROR Executor: Exception in task 0.0 in stage 5.0 (TID 6)
java.lang.ClassCastException: java.lang.Integer cannot be cast to java.sql.Date
at org.apache.spark.sql.execution.vectorized.ColumnVectorUtils.populate(ColumnVectorUtils.java:89)
at org.apache.spark.sql.execution.datasources.parquet.VectorizedParquetRecordReader.initBatch(VectorizedParquetRecordReader.java:185)
at org.apache.spark.sql.execution.datasources.parquet.VectorizedParquetRecordReader.initBatch(VectorizedParquetRecordReader.java:204)
at org.apache.spark.sql.execution.datasources.parquet.ParquetFileFormat$$anonfun$buildReader$1.apply(ParquetFileFormat.scala:362)
...
```
## How was this patch tested?
Unit tests in `SQLQuerySuite`.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#14919 from HyukjinKwon/SPARK-17354.
The secrets leveldb isn't being moved if you run spark shuffle services without yarn nm recovery on and then turn it on. This fixes that. I unfortunately missed this when I ported the patch from our internal branch 2 to master branch due to the changes for the recovery path. Note this only applies to master since it is the only place the yarn nm recovery dir is used.
Unit tests ran and tested on 8 node cluster. Fresh startup with NM recovery, fresh startup no nm recovery, switching between no nm recovery and recovery. Also tested running applications to make sure wasn't affected by rolling upgrade.
Author: Thomas Graves <tgraves@prevailsail.corp.gq1.yahoo.com>
Author: Tom Graves <tgraves@apache.org>
Closes#14999 from tgravescs/SPARK-17433.
