## What changes were proposed in this pull request?
Currently, Spark writes Spark version number into Hive Table properties with `spark.sql.create.version`.
```
parameters:{
spark.sql.sources.schema.part.0={
"type":"struct",
"fields":[{"name":"a","type":"integer","nullable":true,"metadata":{}}]
},
transient_lastDdlTime=1541142761,
spark.sql.sources.schema.numParts=1,
spark.sql.create.version=2.4.0
}
```
This PR aims to write Spark versions to ORC/Parquet file metadata with `org.apache.spark.sql.create.version` because we used `org.apache.` prefix in Parquet metadata already. It's different from Hive Table property key `spark.sql.create.version`, but it seems that we cannot change Hive Table property for backward compatibility.
After this PR, ORC and Parquet file generated by Spark will have the following metadata.
**ORC (`native` and `hive` implmentation)**
```
$ orc-tools meta /tmp/o
File Version: 0.12 with ...
...
User Metadata:
org.apache.spark.sql.create.version=3.0.0
```
**PARQUET**
```
$ parquet-tools meta /tmp/p
...
creator: parquet-mr version 1.10.0 (build 031a6654009e3b82020012a18434c582bd74c73a)
extra: org.apache.spark.sql.create.version = 3.0.0
extra: org.apache.spark.sql.parquet.row.metadata = {"type":"struct","fields":[{"name":"id","type":"long","nullable":false,"metadata":{}}]}
```
## How was this patch tested?
Pass the Jenkins with newly added test cases.
This closes#22255.
Closes#22932 from dongjoon-hyun/SPARK-25102.
Authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
Every time a task is unschedulable because of the condition where no. of task failures < no. of executors available, we currently abort the taskSet - failing the job. This change tries to acquire new executors so that we can complete the job successfully. We try to acquire a new executor only when we can kill an existing idle executor. We fallback to the older implementation where we abort the job if we cannot find an idle executor.
## How was this patch tested?
I performed some manual tests to check and validate the behavior.
```scala
val rdd = sc.parallelize(Seq(1 to 10), 3)
import org.apache.spark.TaskContext
val mapped = rdd.mapPartitionsWithIndex ( (index, iterator) => { if (index == 2) { Thread.sleep(30 * 1000); val attemptNum = TaskContext.get.attemptNumber; if (attemptNum < 3) throw new Exception("Fail for blacklisting")}; iterator.toList.map (x => x + " -> " + index).iterator } )
mapped.collect
```
Closes#22288 from dhruve/bug/SPARK-22148.
Lead-authored-by: Dhruve Ashar <dhruveashar@gmail.com>
Co-authored-by: Dhruve Ashar <dhruve@users.noreply.github.com>
Co-authored-by: Tom Graves <tgraves@apache.org>
Signed-off-by: Thomas Graves <tgraves@apache.org>
## What changes were proposed in this pull request?
Currently definitions of config entries in `core` module are in several files separately. We should move them into `internal/config` to be easy to manage.
## How was this patch tested?
Existing tests.
Closes#22928 from ueshin/issues/SPARK-25926/single_config_file.
Authored-by: Takuya UESHIN <ueshin@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Fix typos and misspellings, per https://github.com/apache/spark-website/pull/158#issuecomment-435790366
## How was this patch tested?
Existing tests.
Closes#22950 from srowen/Typos.
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 saw CoarseGrainedSchedulerBackendSuite failed in my PR and finally reproduced the following error on a very busy machine:
```
sbt.ForkMain$ForkError: org.scalatest.exceptions.TestFailedDueToTimeoutException: The code passed to eventually never returned normally. Attempted 400 times over 10.009828643999999 seconds. Last failure message: ArrayBuffer("2", "0", "3") had length 3 instead of expected length 4.
```
The logs in this test shows executor 1 was not up when the test failed.
```
18/10/30 11:34:03.563 dispatcher-event-loop-12 INFO CoarseGrainedSchedulerBackend$DriverEndpoint: Registered executor NettyRpcEndpointRef(spark-client://Executor) (172.17.0.2:43656) with ID 2
18/10/30 11:34:03.593 dispatcher-event-loop-3 INFO CoarseGrainedSchedulerBackend$DriverEndpoint: Registered executor NettyRpcEndpointRef(spark-client://Executor) (172.17.0.2:43658) with ID 3
18/10/30 11:34:03.629 dispatcher-event-loop-6 INFO CoarseGrainedSchedulerBackend$DriverEndpoint: Registered executor NettyRpcEndpointRef(spark-client://Executor) (172.17.0.2:43654) with ID 0
18/10/30 11:34:03.885 pool-1-thread-1-ScalaTest-running-CoarseGrainedSchedulerBackendSuite INFO CoarseGrainedSchedulerBackendSuite:
===== FINISHED o.a.s.scheduler.CoarseGrainedSchedulerBackendSuite: 'compute max number of concurrent tasks can be launched' =====
```
And the following logs in executor 1 shows it was still doing the initialization when the timeout happened (at 18/10/30 11:34:03.885).
```
18/10/30 11:34:03.463 netty-rpc-connection-0 INFO TransportClientFactory: Successfully created connection to 54b6b6217301/172.17.0.2:33741 after 37 ms (0 ms spent in bootstraps)
18/10/30 11:34:03.959 main INFO DiskBlockManager: Created local directory at /home/jenkins/workspace/core/target/tmp/spark-383518bc-53bd-4d9c-885b-d881f03875bf/executor-61c406e4-178f-40a6-ac2c-7314ee6fb142/blockmgr-03fb84a1-eedc-4055-8743-682eb3ac5c67
18/10/30 11:34:03.993 main INFO MemoryStore: MemoryStore started with capacity 546.3 MB
```
Hence, I think our current 10 seconds is not enough on a slow Jenkins machine. This PR just increases the timeout from 10 seconds to 60 seconds to make the test more stable.
## How was this patch tested?
Jenkins
Closes#22910 from zsxwing/fix-flaky-test.
Authored-by: Shixiong Zhu <zsxwing@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
This avoids having two classes to deal with tokens; now the above
class is a one-stop shop for dealing with delegation tokens. The
YARN backend extends that class instead of doing composition like
before, resulting in a bit less code there too.
The renewer functionality is basically the same code that used to
be in YARN's AMCredentialRenewer. That is also the reason why the
public API of HadoopDelegationTokenManager is a little bit odd;
the YARN AM has some odd requirements for how this all should be
initialized, and the weirdness is needed currently to support that.
Tested:
- YARN with stress app for DT renewal
- Mesos and K8S with basic kerberos tests (both tgt and keytab)
Closes#22624 from vanzin/SPARK-23781.
Authored-by: Marcelo Vanzin <vanzin@cloudera.com>
Signed-off-by: Imran Rashid <irashid@cloudera.com>
## What changes were proposed in this pull request?
When a job finishes, there may be some zombie tasks still running due to stage retry. Since a result stage will never be used by other jobs, running these tasks are just wasting the cluster resource. This PR just asks TaskScheduler to cancel the running tasks of a result stage when it's already finished. Credits go to srinathshankar who suggested this idea to me.
This PR also fixes two minor issues while I'm touching DAGScheduler:
- Invalid spark.job.interruptOnCancel should not crash DAGScheduler.
- Non fatal errors should not crash DAGScheduler.
## How was this patch tested?
The new unit tests.
Closes#22771 from zsxwing/SPARK-25773.
Lead-authored-by: Shixiong Zhu <zsxwing@gmail.com>
Co-authored-by: Shixiong Zhu <shixiong@databricks.com>
Signed-off-by: Shixiong Zhu <zsxwing@gmail.com>
## What changes were proposed in this pull request?
Set main args correctly in BenchmarkBase, to make it accessible for its subclass.
It will benefit:
- BuiltInDataSourceWriteBenchmark
- AvroWriteBenchmark
## How was this patch tested?
manual tests
Closes#22872 from yucai/main_args.
Authored-by: yucai <yyu1@ebay.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
`hsync` has been added as part of SPARK-19531 to get the latest data in the history sever ui, but that is causing the performance overhead and also leading to drop many history log events. `hsync` uses the force `FileChannel.force` to sync the data to the disk and happens for the data pipeline, it is costly operation and making the application to face overhead and drop the events.
I think getting the latest data in history server can be done in different way (no impact to application while writing events), there is an api `DFSInputStream.getFileLength()` which gives the file length including the `lastBlockBeingWrittenLength`(different from `FileStatus.getLen()`), this api can be used when the file status length and previously cached length are equal to verify whether any new data has been written or not, if there is any update in data length then the history server can update the in progress history log. And also I made this change as configurable with the default value false, and can be enabled for history server if users want to see the updated data in ui.
## How was this patch tested?
Added new test and verified manually, with the added conf `spark.history.fs.inProgressAbsoluteLengthCheck.enabled=true`, history server is reading the logs including the last block data which is being written and updating the Web UI with the latest data.
Closes#22752 from devaraj-kavali/SPARK-24787.
Authored-by: Devaraj K <devaraj@apache.org>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
In standalone cluster mode, one could launch driver with supervise mode
enabled. StandaloneRestServer class uses the host and port of current
master as the spark.master property while launching the driver
(even if you are running in HA mode). This class also ignores the
spark.master property passed as part of the request.
Due to the above problem, if the Spark masters switch due to some reason
and your driver is killed unexpectedly and relaunched, it will try to
connect to the master which is in the driver command specified as
-Dspark.master. But this master will be in STANDBY mode and after trying
multiple times, the SparkContext will kill itself (even though secondary
master was alive and healthy).
This change picks the spark.master property from request and uses it to
launch the driver process. Due to this, the driver process has both
masters in -Dspark.master property. Even if the masters switch, SparkContext
can still connect to the ALIVE master and work correctly.
## How was this patch tested?
This patch was manually tested on a standalone cluster running 2.2.1. It was rebased on current master and all tests were executed. I have added a unit test for this change (but since I am new I hope I have covered all).
Closes#21816 from bsikander/rest_driver_fix.
Authored-by: Behroz Sikander <behroz.sikander@sap.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Before the code changes, I tried to run it with 8G memory:
```
build/sbt -mem 8000 "core/testOnly org.apache.spark.serializer.KryoBenchmark"
```
Still I got got OOM.
This is because the lengths of the arrays are random
669ade3a8e/core/src/test/scala/org/apache/spark/serializer/KryoBenchmark.scala (L90-L91)
And the 2D array is usually large: `10000 * Random.nextInt(0, 10000)`
This PR is to fix it and refactor it to use main method.
The benchmark result is also reason compared to the original one.
## How was this patch tested?
Run with
```
bin/spark-submit --class org.apache.spark.serializer.KryoBenchmark core/target/scala-2.11/spark-core_2.11-3.0.0-SNAPSHOT-tests.jar
```
and
```
SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "core/test:runMain org.apache.spark.serializer.KryoBenchmark"
Closes#22663 from gengliangwang/kyroBenchmark.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Remove JavaSparkContextVarargsWorkaround
## How was this patch tested?
Existing tests.
Closes#22729 from srowen/SPARK-25737.
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?
Remove deprecated accumulator v1
## How was this patch tested?
Existing tests.
Closes#22730 from srowen/SPARK-16775.
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?
This is a follow up of #21601, `StreamFileInputFormat` and `WholeTextFileInputFormat` have the same problem.
`Minimum split size pernode 5123456 cannot be larger than maximum split size 4194304
java.io.IOException: Minimum split size pernode 5123456 cannot be larger than maximum split size 4194304
at org.apache.hadoop.mapreduce.lib.input.CombineFileInputFormat.getSplits(CombineFileInputFormat.java: 201)
at org.apache.spark.rdd.BinaryFileRDD.getPartitions(BinaryFileRDD.scala:52)
at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:254)
at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:252)
at scala.Option.getOrElse(Option.scala:121)
at org.apache.spark.rdd.RDD.partitions(RDD.scala:252)
at org.apache.spark.SparkContext.runJob(SparkContext.scala:2138)`
## How was this patch tested?
Added a unit test
Closes#22725 from 10110346/maxSplitSize_node_rack.
Authored-by: liuxian <liu.xian3@zte.com.cn>
Signed-off-by: Thomas Graves <tgraves@apache.org>
## What changes were proposed in this pull request?
Currently, if we try run
```
./start-history-server.sh -h
```
We will get such error
```
java.io.FileNotFoundException: File -h does not exist
```
1. This is not User-Friendly. For option `-h` or `--help`, it should be parsed correctly and show the usage of the class/script.
2. We can remove deprecated options for setting event log directory through command line options.
After fix, we can get following output:
```
Usage: ./sbin/start-history-server.sh [options]
Options:
--properties-file FILE Path to a custom Spark properties file.
Default is conf/spark-defaults.conf.
Configuration options can be set by setting the corresponding JVM system property.
History Server options are always available; additional options depend on the provider.
History Server options:
spark.history.ui.port Port where server will listen for connections
(default 18080)
spark.history.acls.enable Whether to enable view acls for all applications
(default false)
spark.history.provider Name of history provider class (defaults to
file system-based provider)
spark.history.retainedApplications Max number of application UIs to keep loaded in memory
(default 50)
FsHistoryProvider options:
spark.history.fs.logDirectory Directory where app logs are stored
(default: file:/tmp/spark-events)
spark.history.fs.updateInterval How often to reload log data from storage
(in seconds, default: 10)
```
## How was this patch tested?
Manual test
Closes#22699 from gengliangwang/refactorSHSUsage.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
Refactor `JoinBenchmark` to use main method.
1. use `spark-submit`:
```console
bin/spark-submit --class org.apache.spark.sql.execution.benchmark.JoinBenchmark --jars ./core/target/spark-core_2.11-3.0.0-SNAPSHOT-tests.jar ./sql/catalyst/target/spark-sql_2.11-3.0.0-SNAPSHOT-tests.jar
```
2. Generate benchmark result:
```console
SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "sql/test:runMain org.apache.spark.sql.execution.benchmark.JoinBenchmark"
```
## How was this patch tested?
manual tests
Closes#22661 from wangyum/SPARK-25664.
Lead-authored-by: Yuming Wang <yumwang@ebay.com>
Co-authored-by: Yuming Wang <wgyumg@gmail.com>
Co-authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
Cause of the error is wrapped with SparkException, now finding the cause from the wrapped exception and throwing the cause instead of the wrapped exception.
## How was this patch tested?
Verified it manually by checking the cause of the error, it gives the error as shown below.
### Without the PR change
```
[apache-spark]$ ./bin/spark-submit --verbose --master spark://******
....
Error: Exception thrown in awaitResult:
Run with --help for usage help or --verbose for debug output
```
### With the PR change
```
[apache-spark]$ ./bin/spark-submit --verbose --master spark://******
....
Exception in thread "main" org.apache.spark.SparkException: Exception thrown in awaitResult:
at org.apache.spark.util.ThreadUtils$.awaitResult(ThreadUtils.scala:226)
at org.apache.spark.rpc.RpcTimeout.awaitResult(RpcTimeout.scala:75)
....
at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)
Caused by: java.io.IOException: Failed to connect to devaraj-pc1/10.3.66.65:7077
at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:245)
....
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: io.netty.channel.AbstractChannel$AnnotatedConnectException: Connection refused: devaraj-pc1/10.3.66.65:7077
at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
....
at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:138)
... 1 more
Caused by: java.net.ConnectException: Connection refused
... 11 more
```
Closes#22623 from devaraj-kavali/SPARK-25636.
Authored-by: Devaraj K <devaraj@apache.org>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
The commons-crypto library does some questionable error handling internally,
which can lead to JVM crashes if some call into native code fails and cleans
up state it should not.
While the library is not fixed, this change adds some workarounds in Spark code
so that when an error is detected in the commons-crypto side, Spark avoids
calling into the library further.
Tested with existing and added unit tests.
Closes#22557 from vanzin/SPARK-25535.
Authored-by: Marcelo Vanzin <vanzin@cloudera.com>
Signed-off-by: Imran Rashid <irashid@cloudera.com>
While working on another PR, I noticed that there is quite some legacy Java in there that can be beautified. For example the use of features from Java8, such as:
- Collection libraries
- Try-with-resource blocks
No logic has been changed. I think it is important to have a solid codebase with examples that will inspire next PR's to follow up on the best practices.
What are your thoughts on this?
This makes code easier to read, and using try-with-resource makes is less likely to forget to close something.
## What changes were proposed in this pull request?
No changes in the logic of Spark, but more in the aesthetics of the code.
## How was this patch tested?
Using the existing unit tests. Since no logic is changed, the existing unit tests should pass.
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#22637 from Fokko/SPARK-25408.
Authored-by: Fokko Driesprong <fokkodriesprong@godatadriven.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
While working on another PR, I noticed that there is quite some legacy Java in there that can be beautified. For example the use og features from Java8, such as:
- Collection libraries
- Try-with-resource blocks
No code has been changed
What are your thoughts on this?
This makes code easier to read, and using try-with-resource makes is less likely to forget to close something.
## What changes were proposed in this pull request?
(Please fill in changes proposed in this fix)
## 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#22399 from Fokko/SPARK-25408.
Authored-by: Fokko Driesprong <fokkodriesprong@godatadriven.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Rename method `benchmark` in `BenchmarkBase` as `runBenchmarkSuite `. Also add comments.
Currently the method name `benchmark` is a bit confusing. Also the name is the same as instances of `Benchmark`:
f246813afb/sql/hive/src/test/scala/org/apache/spark/sql/hive/orc/OrcReadBenchmark.scala (L330-L339)
## How was this patch tested?
Unit test.
Closes#22599 from gengliangwang/renameBenchmarkSuite.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
This PR adds a rule to force `.toLowerCase(Locale.ROOT)` or `toUpperCase(Locale.ROOT)`.
It produces an error as below:
```
[error] Are you sure that you want to use toUpperCase or toLowerCase without the root locale? In most cases, you
[error] should use toUpperCase(Locale.ROOT) or toLowerCase(Locale.ROOT) instead.
[error] If you must use toUpperCase or toLowerCase without the root locale, wrap the code block with
[error] // scalastyle:off caselocale
[error] .toUpperCase
[error] .toLowerCase
[error] // scalastyle:on caselocale
```
This PR excludes the cases above for SQL code path for external calls like table name, column name and etc.
For test suites, or when it's clear there's no locale problem like Turkish locale problem, it uses `Locale.ROOT`.
One minor problem is, `UTF8String` has both methods, `toLowerCase` and `toUpperCase`, and the new rule detects them as well. They are ignored.
## How was this patch tested?
Manually tested, and Jenkins tests.
Closes#22581 from HyukjinKwon/SPARK-25565.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Since we don't fail a job when `AccumulatorV2.merge` fails, we should try to update the remaining accumulators so that they can still report correct values.
## How was this patch tested?
The new unit test.
Closes#22586 from zsxwing/SPARK-25568.
Authored-by: Shixiong Zhu <zsxwing@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
Heartbeat shouldn't include accumulators for zero metrics.
Heartbeats sent from executors to the driver every 10 seconds contain metrics and are generally on the order of a few KBs. However, for large jobs with lots of tasks, heartbeats can be on the order of tens of MBs, causing tasks to die with heartbeat failures. We can mitigate this by not sending zero metrics to the driver.
## 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#22473 from mukulmurthy/25449-heartbeat.
Authored-by: Mukul Murthy <mukul.murthy@gmail.com>
Signed-off-by: Shixiong Zhu <zsxwing@gmail.com>
## What changes were proposed in this pull request?
The specified test in OpenHashMapSuite to test large items is somehow flaky to throw OOM.
By considering the original work #6763 that added this test, the test can be against OpenHashSetSuite. And by doing this should be to save memory because OpenHashMap allocates two more arrays when growing the map/set.
## How was this patch tested?
Existing tests.
Closes#22569 from viirya/SPARK-25542.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
Special case the situation where we know the partioner and the number of requested partions output is the same as the current partioner to avoid a shuffle and instead compute distinct inside of each partion.
## How was this patch tested?
New unit test that verifies partitioner does not change if the partitioner is known and distinct is called with the same target # of partition.
Closes#22010 from holdenk/SPARK-21436-take-advantage-of-known-partioner-for-distinct-on-rdds.
Authored-by: Holden Karau <holden@pigscanfly.ca>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Previously SPARK-24519 created a modifiable config SHUFFLE_MIN_NUM_PARTS_TO_HIGHLY_COMPRESS. However, the config is being parsed for every creation of MapStatus, which could be very expensive. Another problem with the previous approach is that it created the illusion that this can be changed dynamically at runtime, which was not true. This PR changes it so the config is computed only once.
## How was this patch tested?
Removed a test case that's no longer valid.
Closes#22521 from rxin/SPARK-24519.
Authored-by: Reynold Xin <rxin@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
Currently there are two classes with the same naming BenchmarkBase:
1. `org.apache.spark.util.BenchmarkBase`
2. `org.apache.spark.sql.execution.benchmark.BenchmarkBase`
This is very confusing. And the benchmark object `org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark` is using the one in `org.apache.spark.util.BenchmarkBase`, while there is another class `BenchmarkBase` in the same package of it...
Here I propose:
1. the package `org.apache.spark.util.BenchmarkBase` should be in test package of core module. Move it to package `org.apache.spark.benchmark` .
2. Move `org.apache.spark.util.Benchmark` to test package of core module. Move it to package `org.apache.spark.benchmark` .
3. Rename the class `org.apache.spark.sql.execution.benchmark.BenchmarkBase` as `BenchmarkWithCodegen`
## How was this patch tested?
Unit test
Closes#22513 from gengliangwang/refactorBenchmarkBase.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
A continuation of squito's executor plugin task. By his request I took his code and added testing and moved the plugin initialization to a separate thread.
Executor plugins now run on one separate thread, so the executor does not wait on them. Added testing.
## How was this patch tested?
Added test cases that test using a sample plugin.
Closes#22192 from NiharS/executorPlugin.
Lead-authored-by: Nihar Sheth <niharrsheth@gmail.com>
Co-authored-by: NiharS <niharrsheth@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
This goes to revert sequential PRs based on some discussion and comments at https://github.com/apache/spark/pull/16677#issuecomment-422650759.
#22344#22330#22239#16677
## How was this patch tested?
Existing tests.
Closes#22481 from viirya/revert-SPARK-19355-1.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
We've seen some flakiness in jenkins in SchedulerIntegrationSuite which looks like it just needs a
longer timeout.
Closes#22385 from squito/SPARK-25400.
Authored-by: Imran Rashid <irashid@cloudera.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
This PR ensures to call `super.afterAll()` in `override afterAll()` method for test suites.
* Some suites did not call `super.afterAll()`
* Some suites may call `super.afterAll()` only under certain condition
* Others never call `super.afterAll()`.
This PR also ensures to call `super.beforeAll()` in `override beforeAll()` for test suites.
## How was this patch tested?
Existing UTs
Closes#22337 from kiszk/SPARK-25338.
Authored-by: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
Correct some comparisons between unrelated types to what they seem to… have been trying to do
## How was this patch tested?
Existing tests.
Closes#22384 from srowen/SPARK-25398.
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?
We will update block info coming from executors, at the timing like caching a RDD. However, when removing RDDs with unpersisting, we don't ask to update block info. So the block info is not updated.
We can fix this with few options:
1. Ask to update block info when unpersisting
This is simplest but changes driver-executor communication a bit.
2. Update block info when processing the event of unpersisting RDD
We send a `SparkListenerUnpersistRDD` event when unpersisting RDD. When processing this event, we can update block info of the RDD. This only changes event processing code so the risk seems to be lower.
Currently this patch takes option 2 for lower risk. If we agree first option has no risk, we can change to it.
## How was this patch tested?
Unit tests.
Closes#22341 from viirya/SPARK-24889.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
Stop trimming values of properties loaded from a file
## How was this patch tested?
Added unit test demonstrating the issue hit in production.
Closes#22213 from gerashegalov/gera/SPARK-25221.
Authored-by: Gera Shegalov <gera@apache.org>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
When running TPC-DS benchmarks on 2.4 release, npoggi and winglungngai saw more than 10% performance regression on the following queries: q67, q24a and q24b. After we applying the PR https://github.com/apache/spark/pull/22338, the performance regression still exists. If we revert the changes in https://github.com/apache/spark/pull/19222, npoggi and winglungngai found the performance regression was resolved. Thus, this PR is to revert the related changes for unblocking the 2.4 release.
In the future release, we still can continue the investigation and find out the root cause of the regression.
## How was this patch tested?
The existing test cases
Closes#22361 from gatorsmile/revertMemoryBlock.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
Add new executor level memory metrics (JVM used memory, on/off heap execution memory, on/off heap storage memory, on/off heap unified memory, direct memory, and mapped memory), and expose via the executors REST API. This information will help provide insight into how executor and driver JVM memory is used, and for the different memory regions. It can be used to help determine good values for spark.executor.memory, spark.driver.memory, spark.memory.fraction, and spark.memory.storageFraction.
## What changes were proposed in this pull request?
An ExecutorMetrics class is added, with jvmUsedHeapMemory, jvmUsedNonHeapMemory, onHeapExecutionMemory, offHeapExecutionMemory, onHeapStorageMemory, and offHeapStorageMemory, onHeapUnifiedMemory, offHeapUnifiedMemory, directMemory and mappedMemory. The new ExecutorMetrics is sent by executors to the driver as part of the Heartbeat. A heartbeat is added for the driver as well, to collect these metrics for the driver.
The EventLoggingListener store information about the peak values for each metric, per active stage and executor. When a StageCompleted event is seen, a StageExecutorsMetrics event will be logged for each executor, with peak values for the stage.
The AppStatusListener records the peak values for each memory metric.
The new memory metrics are added to the executors REST API.
## How was this patch tested?
New unit tests have been added. This was also tested on our cluster.
Author: Edwina Lu <edlu@linkedin.com>
Author: Imran Rashid <irashid@cloudera.com>
Author: edwinalu <edwina.lu@gmail.com>
Closes#21221 from edwinalu/SPARK-23429.2.
## What changes were proposed in this pull request?
This adds a test following https://github.com/apache/spark/pull/21638
## How was this patch tested?
Existing tests and new test.
Closes#22356 from srowen/SPARK-22357.2.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
This is not a perfect solution. It is designed to minimize complexity on the basis of solving problems.
It is effective for English, Chinese characters, Japanese, Korean and so on.
```scala
before:
+---+---------------------------+-------------+
|id |中国 |s2 |
+---+---------------------------+-------------+
|1 |ab |[a] |
|2 |null |[中国, abc] |
|3 |ab1 |[hello world]|
|4 |か行 きゃ(kya) きゅ(kyu) きょ(kyo) |[“中国] |
|5 |中国（你好）a |[“中（国）, 312] |
|6 |中国山(东)服务区 |[“中(国）] |
|7 |中国山东服务区 |[中(国)] |
|8 | |[中国] |
+---+---------------------------+-------------+
after:
+---+-----------------------------------+----------------+
|id |中国 |s2 |
+---+-----------------------------------+----------------+
|1 |ab |[a] |
|2 |null |[中国, abc] |
|3 |ab1 |[hello world] |
|4 |か行 きゃ(kya) きゅ(kyu) きょ(kyo) |[“中国] |
|5 |中国（你好）a |[“中（国）, 312]|
|6 |中国山(东)服务区 |[“中(国）] |
|7 |中国山东服务区 |[中(国)] |
|8 | |[中国] |
+---+-----------------------------------+----------------+
```
## What changes were proposed in this pull request?
When there are wide characters such as Chinese characters or Japanese characters in the data, the show method has a alignment problem.
Try to fix this problem.
## How was this patch tested?
(Please explain how this patch was tested. E.g. unit tests, integration tests, manual tests)
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#22048 from xuejianbest/master.
Authored-by: xuejianbest <384329882@qq.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Upgrade chill to 0.9.3, Kryo to 4.0.2, to get bug fixes and improvements.
The resolved tickets includes:
- SPARK-25258 Upgrade kryo package to version 4.0.2
- SPARK-23131 Kryo raises StackOverflow during serializing GLR model
- SPARK-25176 Kryo fails to serialize a parametrised type hierarchy
More details:
https://github.com/twitter/chill/releases/tag/v0.9.3cc3910d501
## How was this patch tested?
Existing tests.
Closes#22179 from wangyum/SPARK-23131.
Lead-authored-by: Yuming Wang <yumwang@ebay.com>
Co-authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
An alternative fix for https://github.com/apache/spark/pull/21698
When Spark rerun tasks for an RDD, there are 3 different behaviors:
1. determinate. Always return the same result with same order when rerun.
2. unordered. Returns same data set in random order when rerun.
3. indeterminate. Returns different result when rerun.
Normally Spark doesn't need to care about it. Spark runs stages one by one, when a task is failed, just rerun it. Although the rerun task may return a different result, users will not be surprised.
However, Spark may rerun a finished stage when seeing fetch failures. When this happens, Spark needs to rerun all the tasks of all the succeeding stages if the RDD output is indeterminate, because the input of the succeeding stages has been changed.
If the RDD output is determinate, we only need to rerun the failed tasks of the succeeding stages, because the input doesn't change.
If the RDD output is unordered, it's same as determinate, because shuffle partitioner is always deterministic(round-robin partitioner is not a shuffle partitioner that extends `org.apache.spark.Partitioner`), so the reducers will still get the same input data set.
This PR fixed the failure handling for `repartition`, to avoid correctness issues.
For `repartition`, it applies a stateful map function to generate a round-robin id, which is order sensitive and makes the RDD's output indeterminate. When the stage contains `repartition` reruns, we must also rerun all the tasks of all the succeeding stages.
**future improvement:**
1. Currently we can't rollback and rerun a shuffle map stage, and just fail. We should fix it later. https://issues.apache.org/jira/browse/SPARK-25341
2. Currently we can't rollback and rerun a result stage, and just fail. We should fix it later. https://issues.apache.org/jira/browse/SPARK-25342
3. We should provide public API to allow users to tag the random level of the RDD's computing function.
## How is this pull request tested?
a new test case
Closes#22112 from cloud-fan/repartition.
Lead-authored-by: Wenchen Fan <wenchen@databricks.com>
Co-authored-by: Xingbo Jiang <xingbo.jiang@databricks.com>
Signed-off-by: Xiao Li <gatorsmile@gmail.com>
Running a large Spark job with speculation turned on was causing executor heartbeats to time out on the driver end after sometime and eventually, after hitting the max number of executor failures, the job would fail.
## What changes were proposed in this pull request?
The main reason for the heartbeat timeouts was that the heartbeat-receiver-event-loop-thread was blocked waiting on the TaskSchedulerImpl object which was being held by one of the dispatcher-event-loop threads executing the method dequeueSpeculativeTasks() in TaskSetManager.scala. On further analysis of the heartbeat receiver method executorHeartbeatReceived() in TaskSchedulerImpl class, we found out that instead of waiting to acquire the lock on the TaskSchedulerImpl object, we can remove that lock and make the operations to the global variables inside the code block to be atomic. The block of code in that method only uses one global HashMap taskIdToTaskSetManager. Making that map a ConcurrentHashMap, we are ensuring atomicity of operations and speeding up the heartbeat receiver thread operation.
## How was this patch tested?
Screenshots of the thread dump have been attached below:
**heartbeat-receiver-event-loop-thread:**
<img width="1409" alt="screen shot 2018-08-24 at 9 19 57 am" src="https://user-images.githubusercontent.com/22228190/44593413-e25df780-a788-11e8-9520-176a18401a59.png">
**dispatcher-event-loop-thread:**
<img width="1409" alt="screen shot 2018-08-24 at 9 21 56 am" src="https://user-images.githubusercontent.com/22228190/44593484-13d6c300-a789-11e8-8d88-34b1d51d4541.png">
Closes#22221 from pgandhi999/SPARK-25231.
Authored-by: pgandhi <pgandhi@oath.com>
Signed-off-by: Thomas Graves <tgraves@apache.org>
The problem occurs because stage object is removed from liveStages in
AppStatusListener onStageCompletion. Because of this any onTaskEnd event
received after onStageCompletion event do not update stage metrics.
The fix is to retain stage objects in liveStages until all tasks are complete.
1. Fixed the reproducible example posted in the JIRA
2. Added unit test
Closes#22209 from ankuriitg/ankurgupta/SPARK-24415.
Authored-by: ankurgupta <ankur.gupta@cloudera.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
The configuration parameter "spark.shuffle.service.enabled" has defined in `package.scala`, and it is also used in many place, so we can replace it with `SHUFFLE_SERVICE_ENABLED`.
and unified this configuration parameter "spark.shuffle.service.port" together.
## How was this patch tested?
N/A
Closes#22306 from 10110346/unifiedserviceenable.
Authored-by: liuxian <liu.xian3@zte.com.cn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
