### What changes were proposed in this pull request?
`ExternalBlockHandler` exposes 4 metrics which are Dropwizard `Timer` metrics, and are named with a `millis` suffix:
```
private final Timer openBlockRequestLatencyMillis = new Timer();
private final Timer registerExecutorRequestLatencyMillis = new Timer();
private final Timer fetchMergedBlocksMetaLatencyMillis = new Timer();
private final Timer finalizeShuffleMergeLatencyMillis = new Timer();
```
However these Dropwizard Timers by default use nanoseconds ([documentation](https://metrics.dropwizard.io/3.2.3/getting-started.html#timers)).
This causes `YarnShuffleServiceMetrics` to expose confusingly-named metrics like `openBlockRequestLatencyMillis_nanos_max` (the actual values are currently in nanos).
This PR adds a new `Timer` subclass, `TimerWithCustomTimeUnit`, which accepts a `TimeUnit` at creation time and exposes timing information using this time unit when values are read. Internally, values are still stored with nanosecond-level precision. The `Timer` metrics within `ExternalBlockHandler` are updated to use the new class with milliseconds as the unit. The logic to include the `nanos` suffix in the metric name within `YarnShuffleServiceMetrics` has also been removed, with the assumption that the metric name itself includes the units.
### Does this PR introduce _any_ user-facing change?
Yes, there are two changes.
First, the names for metrics exposed by `ExternalBlockHandler` via `YarnShuffleServiceMetrics` such as `openBlockRequestLatencyMillis_nanos_max` and `openBlockRequestLatencyMillis_nanos_50thPercentile` have been changed to remove the `_nanos` suffix. This would be considered a breaking change, but these names were only exposed as part of #32388, which has not yet been released (slated for 3.2.0). New names are like `openBlockRequestLatencyMillis_max` and `openBlockRequestLatencyMillis_50thPercentile`
Second, the values of the metrics themselves have changed, to expose milliseconds instead of nanoseconds. Note that this does not affect metrics such as `openBlockRequestLatencyMillis_count` or `openBlockRequestLatencyMillis_rate1`, only the `Snapshot`-related metrics (`max`, `median`, percentiles, etc.). For the YARN case, these metrics were also introduced by #32388, and thus also have not yet been released. It was possible for the nanosecond values to be consumed by some other metrics reporter reading the Dropwizard metrics directly, but I'm not aware of any such usages.
### How was this patch tested?
Unit tests have been updated.
Closes#33116 from xkrogen/xkrogen-SPARK-35259-ess-fix-metric-unit-prefix.
Authored-by: Erik Krogen <xkrogen@apache.org>
Signed-off-by: yi.wu <yi.wu@databricks.com>
### What changes were proposed in this pull request?
This PR aims to update `master` branch version to 3.3.0-SNAPSHOT.
### Why are the changes needed?
Start to prepare Apache Spark 3.3.0 and the published snapshot version should not conflict with `branch-3.2`.
### Does this PR introduce _any_ user-facing change?
N/A.
### How was this patch tested?
Pass the CIs.
Closes#33196 from dongjoon-hyun/SPARK-35996.
Authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This adds two new additional metrics to `ExternalBlockHandler`:
- `blockTransferRate` -- for indicating the rate of transferring blocks, vs. the data within them
- `blockTransferAvgSize_1min` -- a 1-minute trailing average of block sizes transferred by the ESS
Additionally, this enhances `YarnShuffleServiceMetrics` to expose the histogram/`Snapshot` information from `Timer` metrics within `ExternalBlockHandler`.
### Why are the changes needed?
Currently `ExternalBlockHandler` exposes some useful metrics, but is lacking around metrics for the rate of block transfers. We have `blockTransferRateBytes` to tell us the rate of _bytes_, but no metric to tell us the rate of _blocks_, which is especially relevant when running the ESS on HDDs that are sensitive to random reads. Many small block transfers can have a negative impact on performance, but won't show up as a spike in `blockTransferRateBytes` since the sizes are small. Thus the new metrics to show information around average block size and block transfer rate are very useful to monitor the health/performance of the ESS, especially when running on HDDs.
For the `YarnShuffleServiceMetrics`, currently the three `Timer` metrics exposed by `ExternalBlockHandler` are being underutilized in a YARN-based environment -- they are basically treated as a `Meter`, only exposing rate-based information, when the metrics themselves are collected detailed histograms of timing information. We should expose this information for better observability.
### Does this PR introduce _any_ user-facing change?
Yes, there are two entirely new metrics for the ESS, as documented in `monitoring.md`. Additionally in a YARN environment, `Timer` metrics exposed by the ESS will include more rich timing information.
### How was this patch tested?
New unit tests are added to verify that new metrics are showing up as expected.
We have been running this patch internally for approx. 1 year and have found it to be useful for monitoring the health of ESS and diagnosing performance issues.
Closes#32388 from xkrogen/xkrogen-SPARK-35258-ess-new-metrics.
Authored-by: Erik Krogen <xkrogen@apache.org>
Signed-off-by: Mridul Muralidharan <mridul<at>gmail.com>
### What changes were proposed in this pull request?
Add a new config, `spark.shuffle.service.name`, which allows for Spark applications to look for a YARN shuffle service which is defined at a name other than the default `spark_shuffle`.
Add a new config, `spark.yarn.shuffle.service.metrics.namespace`, which allows for configuring the namespace used when emitting metrics from the shuffle service into the NodeManager's `metrics2` system.
Add a new mechanism by which to override shuffle service configurations independently of the configurations in the NodeManager. When a resource `spark-shuffle-site.xml` is present on the classpath of the shuffle service, the configs present within it will be used to override the configs coming from `yarn-site.xml` (via the NodeManager).
### Why are the changes needed?
There are two use cases which can benefit from these changes.
One use case is to run multiple instances of the shuffle service side-by-side in the same NodeManager. This can be helpful, for example, when running a YARN cluster with a mixed workload of applications running multiple Spark versions, since a given version of the shuffle service is not always compatible with other versions of Spark (e.g. see SPARK-27780). With this PR, it is possible to run two shuffle services like `spark_shuffle` and `spark_shuffle_3.2.0`, one of which is "legacy" and one of which is for new applications. This is possible because YARN versions since 2.9.0 support the ability to run shuffle services within an isolated classloader (see YARN-4577), meaning multiple Spark versions can coexist.
Besides this, the separation of shuffle service configs into `spark-shuffle-site.xml` can be useful for administrators who want to change and/or deploy Spark shuffle service configurations independently of the configurations for the NodeManager (e.g., perhaps they are owned by two different teams).
### Does this PR introduce _any_ user-facing change?
Yes. There are two new configurations related to the external shuffle service, and a new mechanism which can optionally be used to configure the shuffle service. `docs/running-on-yarn.md` has been updated to provide user instructions; please see this guide for more details.
### How was this patch tested?
In addition to the new unit tests added, I have deployed this to a live YARN cluster and successfully deployed two Spark shuffle services simultaneously, one running a modified version of Spark 2.3.0 (which supports some of the newer shuffle protocols) and one running Spark 3.1.1. Spark applications of both versions are able to communicate with their respective shuffle services without issue.
Closes#31936 from xkrogen/xkrogen-SPARK-34828-shufflecompat-config-from-classpath.
Authored-by: Erik Krogen <xkrogen@apache.org>
Signed-off-by: Thomas Graves <tgraves@apache.org>
### What changes were proposed in this pull request?
This:
1. switches Spark to use shaded Hadoop clients, namely hadoop-client-api and hadoop-client-runtime, for Hadoop 3.x.
2. upgrade built-in version for Hadoop 3.x to Hadoop 3.2.2
Note that for Hadoop 2.7, we'll still use the same modules such as hadoop-client.
In order to still keep default Hadoop profile to be hadoop-3.2, this defines the following Maven properties:
```
hadoop-client-api.artifact
hadoop-client-runtime.artifact
hadoop-client-minicluster.artifact
```
which default to:
```
hadoop-client-api
hadoop-client-runtime
hadoop-client-minicluster
```
but all switch to `hadoop-client` when the Hadoop profile is hadoop-2.7. A side affect from this is we'll import the same dependency multiple times. For this I have to disable Maven enforcer `banDuplicatePomDependencyVersions`.
Besides above, there are the following changes:
- explicitly add a few dependencies which are imported via transitive dependencies from Hadoop jars, but are removed from the shaded client jars.
- removed the use of `ProxyUriUtils.getPath` from `ApplicationMaster` which is a server-side/private API.
- modified `IsolatedClientLoader` to exclude `hadoop-auth` jars when Hadoop version is 3.x. This change should only matter when we're not sharing Hadoop classes with Spark (which is _mostly_ used in tests).
### Why are the changes needed?
Hadoop 3.2.2 is released with new features and bug fixes, so it's good for the Spark community to adopt it. However, latest Hadoop versions starting from Hadoop 3.2.1 have upgraded to use Guava 27+. In order to resolve Guava conflicts, this takes the approach by switching to shaded client jars provided by Hadoop. This also has the benefits of avoid pulling other 3rd party dependencies from Hadoop side so as to avoid more potential future conflicts.
### Does this PR introduce _any_ user-facing change?
When people use Spark with `hadoop-provided` option, they should make sure class path contains `hadoop-client-api` and `hadoop-client-runtime` jars. In addition, they may need to make sure these jars appear before other Hadoop jars in the order. Otherwise, classes may be loaded from the other non-shaded Hadoop jars and cause potential conflicts.
### How was this patch tested?
Relying on existing tests.
Closes#30701 from sunchao/test-hadoop-3.2.2.
Authored-by: Chao Sun <sunchao@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This PR aims to update `master` branch version to 3.2.0-SNAPSHOT.
### Why are the changes needed?
Start to prepare Apache Spark 3.2.0.
### Does this PR introduce _any_ user-facing change?
N/A.
### How was this patch tested?
Pass the CIs.
Closes#30606 from dongjoon-hyun/SPARK-3.2.
Authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This reverts commit SPARK-33212 (cb3fa6c936) mostly with three exceptions:
1. `SparkSubmitUtils` was updated recently by SPARK-33580
2. `resource-managers/yarn/pom.xml` was updated recently by SPARK-33104 to add `hadoop-yarn-server-resourcemanager` test dependency.
3. Adjust `com.fasterxml.jackson.module:jackson-module-jaxb-annotations` dependency in K8s module which is updated recently by SPARK-33471.
### Why are the changes needed?
According to [HADOOP-16080](https://issues.apache.org/jira/browse/HADOOP-16080) since Apache Hadoop 3.1.1, `hadoop-aws` doesn't work with `hadoop-client-api`. It fails at write operation like the following.
**1. Spark distribution with `-Phadoop-cloud`**
```scala
$ bin/spark-shell --conf spark.hadoop.fs.s3a.access.key=$AWS_ACCESS_KEY_ID --conf spark.hadoop.fs.s3a.secret.key=$AWS_SECRET_ACCESS_KEY
20/11/30 23:01:24 WARN NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
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 available as 'sc' (master = local[*], app id = local-1606806088715).
Spark session available as 'spark'.
Welcome to
____ __
/ __/__ ___ _____/ /__
_\ \/ _ \/ _ `/ __/ '_/
/___/ .__/\_,_/_/ /_/\_\ version 3.1.0-SNAPSHOT
/_/
Using Scala version 2.12.10 (OpenJDK 64-Bit Server VM, Java 1.8.0_272)
Type in expressions to have them evaluated.
Type :help for more information.
scala> spark.read.parquet("s3a://dongjoon/users.parquet").show
20/11/30 23:01:34 WARN MetricsConfig: Cannot locate configuration: tried hadoop-metrics2-s3a-file-system.properties,hadoop-metrics2.properties
+------+--------------+----------------+
| name|favorite_color|favorite_numbers|
+------+--------------+----------------+
|Alyssa| null| [3, 9, 15, 20]|
| Ben| red| []|
+------+--------------+----------------+
scala> Seq(1).toDF.write.parquet("s3a://dongjoon/out.parquet")
20/11/30 23:02:14 ERROR Executor: Exception in task 0.0 in stage 2.0 (TID 2)/ 1]
java.lang.NoSuchMethodError: org.apache.hadoop.util.SemaphoredDelegatingExecutor.<init>(Lcom/google/common/util/concurrent/ListeningExecutorService;IZ)V
```
**2. Spark distribution without `-Phadoop-cloud`**
```scala
$ bin/spark-shell --conf spark.hadoop.fs.s3a.access.key=$AWS_ACCESS_KEY_ID --conf spark.hadoop.fs.s3a.secret.key=$AWS_SECRET_ACCESS_KEY -c spark.eventLog.enabled=true -c spark.eventLog.dir=s3a://dongjoon/spark-events/ --packages org.apache.hadoop:hadoop-aws:3.2.0,org.apache.hadoop:hadoop-common:3.2.0
...
java.lang.NoSuchMethodError: org.apache.hadoop.util.SemaphoredDelegatingExecutor.<init>(Lcom/google/common/util/concurrent/ListeningExecutorService;IZ)V
at org.apache.hadoop.fs.s3a.S3AFileSystem.create(S3AFileSystem.java:772)
```
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass the CI.
Closes#30508 from dongjoon-hyun/SPARK-33212-REVERT.
Authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This is a follow-up fix for the failing tests in `YarnShuffleServiceSuite.java`. This java class was introduced in https://github.com/apache/spark/pull/30062. The tests in the class fail when run with hadoop-2.7 profile:
```
[ERROR] testCreateDefaultMergedShuffleFileManagerInstance(org.apache.spark.network.yarn.YarnShuffleServiceSuite) Time elapsed: 0.627 s <<< ERROR!
java.lang.NoClassDefFoundError: org/apache/commons/logging/LogFactory
at org.apache.spark.network.yarn.YarnShuffleServiceSuite.testCreateDefaultMergedShuffleFileManagerInstance(YarnShuffleServiceSuite.java:37)
Caused by: java.lang.ClassNotFoundException: org.apache.commons.logging.LogFactory
at org.apache.spark.network.yarn.YarnShuffleServiceSuite.testCreateDefaultMergedShuffleFileManagerInstance(YarnShuffleServiceSuite.java:37)
[ERROR] testCreateRemoteBlockPushResolverInstance(org.apache.spark.network.yarn.YarnShuffleServiceSuite) Time elapsed: 0 s <<< ERROR!
java.lang.NoClassDefFoundError: Could not initialize class org.apache.spark.network.yarn.YarnShuffleService
at org.apache.spark.network.yarn.YarnShuffleServiceSuite.testCreateRemoteBlockPushResolverInstance(YarnShuffleServiceSuite.java:47)
[ERROR] testInvalidClassNameOfMergeManagerWillUseNoOpInstance(org.apache.spark.network.yarn.YarnShuffleServiceSuite) Time elapsed: 0.001 s <<< ERROR!
java.lang.NoClassDefFoundError: Could not initialize class org.apache.spark.network.yarn.YarnShuffleService
at org.apache.spark.network.yarn.YarnShuffleServiceSuite.testInvalidClassNameOfMergeManagerWillUseNoOpInstance(YarnShuffleServiceSuite.java:57)
```
A test suit for `YarnShuffleService` did exist here:
`resource-managers/yarn/src/test/scala/org/apache/spark/network/yarn/YarnShuffleServiceSuite.scala`
I missed this when I created `common/network-yarn/src/test/java/org/apache/spark/network/yarn/YarnShuffleServiceSuite.java`. Moving all the new tests to the earlier test suite fixes the failures with hadoop-2.7 even though why this happened is not clear.
### Why are the changes needed?
The newly added tests are failing when run with hadoop profile 2.7
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Ran the unit tests with the default profile as well as hadoop 2.7 profile.
`build/mvn test -Dtest=none -DwildcardSuites=org.apache.spark.network.yarn.YarnShuffleServiceSuite -Phadoop-2.7 -Pyarn`
```
Run starting. Expected test count is: 11
YarnShuffleServiceSuite:
- executor state kept across NM restart
- removed applications should not be in registered executor file
- shuffle service should be robust to corrupt registered executor file
- get correct recovery path
- moving recovery file from NM local dir to recovery path
- service throws error if cannot start
- recovery db should not be created if NM recovery is not enabled
- SPARK-31646: metrics should be registered into Node Manager's metrics system
- create default merged shuffle file manager instance
- create remote block push resolver instance
- invalid class name of merge manager will use noop instance
Run completed in 2 seconds, 572 milliseconds.
Total number of tests run: 11
Suites: completed 2, aborted 0
Tests: succeeded 11, failed 0, canceled 0, ignored 0, pending 0
All tests passed.
```
Closes#30349 from otterc/SPARK-32916-followup.
Authored-by: Chandni Singh <singh.chandni@gmail.com>
Signed-off-by: Thomas Graves <tgraves@apache.org>
### What changes were proposed in this pull request?
This is one of the patches for SPIP [SPARK-30602](https://issues.apache.org/jira/browse/SPARK-30602) which is needed for push-based shuffle.
Summary of changes:
- Adds an implementation of `MergedShuffleFileManager` which was introduced with [Spark 32915](https://issues.apache.org/jira/browse/SPARK-32915).
- Integrated the push-based shuffle service with `YarnShuffleService`.
### Why are the changes needed?
Refer to the SPIP in [SPARK-30602](https://issues.apache.org/jira/browse/SPARK-30602).
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Added unit tests.
The reference PR with the consolidated changes covering the complete implementation is also provided in [SPARK-30602](https://issues.apache.org/jira/browse/SPARK-30602).
We have already verified the functionality and the improved performance as documented in the SPIP doc.
Lead-authored-by: Min Shen mshenlinkedin.com
Co-authored-by: Chandni Singh chsinghlinkedin.com
Co-authored-by: Ye Zhou yezhoulinkedin.com
Closes#30062 from otterc/SPARK-32916.
Lead-authored-by: Chandni Singh <singh.chandni@gmail.com>
Co-authored-by: Chandni Singh <chsingh@linkedin.com>
Co-authored-by: Ye Zhou <yezhou@linkedin.com>
Co-authored-by: Min Shen <mshen@linkedin.com>
Signed-off-by: Mridul Muralidharan <mridul<at>gmail.com>
### What changes were proposed in this pull request?
This switches Spark to use shaded Hadoop clients, namely hadoop-client-api and hadoop-client-runtime, for Hadoop 3.x. For Hadoop 2.7, we'll still use the same modules such as hadoop-client.
In order to still keep default Hadoop profile to be hadoop-3.2, this defines the following Maven properties:
```
hadoop-client-api.artifact
hadoop-client-runtime.artifact
hadoop-client-minicluster.artifact
```
which default to:
```
hadoop-client-api
hadoop-client-runtime
hadoop-client-minicluster
```
but all switch to `hadoop-client` when the Hadoop profile is hadoop-2.7. A side affect from this is we'll import the same dependency multiple times. For this I have to disable Maven enforcer `banDuplicatePomDependencyVersions`.
Besides above, there are the following changes:
- explicitly add a few dependencies which are imported via transitive dependencies from Hadoop jars, but are removed from the shaded client jars.
- removed the use of `ProxyUriUtils.getPath` from `ApplicationMaster` which is a server-side/private API.
- modified `IsolatedClientLoader` to exclude `hadoop-auth` jars when Hadoop version is 3.x. This change should only matter when we're not sharing Hadoop classes with Spark (which is _mostly_ used in tests).
### Why are the changes needed?
This serves two purposes:
- to unblock Spark from upgrading to Hadoop 3.2.2/3.3.0+. Latest Hadoop versions have upgraded to use Guava 27+ and in order to adopt the latest Hadoop versions in Spark, we'll need to resolve the Guava conflicts. This takes the approach by switching to shaded client jars provided by Hadoop.
- avoid pulling 3rd party dependencies from Hadoop and avoid potential future conflicts.
### Does this PR introduce _any_ user-facing change?
When people use Spark with `hadoop-provided` option, they should make sure class path contains `hadoop-client-api` and `hadoop-client-runtime` jars. In addition, they may need to make sure these jars appear before other Hadoop jars in the order. Otherwise, classes may be loaded from the other non-shaded Hadoop jars and cause potential conflicts.
### How was this patch tested?
Relying on existing tests.
Closes#29843 from sunchao/SPARK-29250.
Authored-by: Chao Sun <sunchao@apple.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>
### What changes were proposed in this pull request?
Register `NettyMemoryMetrics` into Node Manager's metrics system through `YarnShuffleServiceMetrics`.
- usedDirectMemory
- usedHeapMemory
### Why are the changes needed?
Such that `NettyMemoryMetrics` can be exposed through Node Manager's JMX.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Update UT to ensure NettyMemoryMetrics are registered into Node Manager's metrics system.
Closes#28416 from manuzhang/spark-31611.
Authored-by: tianlzhang <tianlzhang@ebay.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Close idle connections at shuffle server side when an `IdleStateEvent` is triggered after `spark.shuffle.io.connectionTimeout` or `spark.network.timeout` time. It's based on following investigations.
1. We found connections on our clusters building up continuously (> 10k for some nodes). Is that normal ? We don't think so.
2. We looked into the connections on one node and found there were a lot of half-open connections. (connections only existed on one node)
3. We also checked those connections were very old (> 21 hours). (FYI, https://superuser.com/questions/565991/how-to-determine-the-socket-connection-up-time-on-linux)
4. Looking at the code, TransportContext registers an IdleStateHandler which should fire an IdleStateEvent when timeout. We did a heap dump of the YarnShuffleService and checked the attributes of IdleStateHandler. It turned out firstAllIdleEvent of many IdleStateHandlers were already false so IdleStateEvent were already fired.
5. Finally, we realized the IdleStateEvent would not be handled since closeIdleConnections are hardcoded to false for YarnShuffleService.
### Why are the changes needed?
Idle connections to YarnShuffleService could never be closed, and will be accumulating and taking up memory and file descriptors.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Existing tests.
Closes#27998 from manuzhang/spark-31219.
Authored-by: manuzhang <owenzhang1990@gmail.com>
Signed-off-by: Thomas Graves <tgraves@apache.org>
### What changes were proposed in this pull request?
This patch is to bump the master branch version to 3.1.0-SNAPSHOT.
### Why are the changes needed?
N/A
### Does this PR introduce any user-facing change?
N/A
### How was this patch tested?
N/A
Closes#27698 from gatorsmile/updateVersion.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Remove usages of Guava that no longer work in Guava 27, and replace with workalikes. I'll comment on key types of changes below.
### Why are the changes needed?
Hadoop 3.2.1 uses Guava 27, so this helps us avoid problems running on Hadoop 3.2.1+ and generally lowers our exposure to Guava.
### Does this PR introduce any user-facing change?
Should not be, but see notes below on hash codes and toString.
### How was this patch tested?
Existing tests will verify whether these changes break anything for Guava 14.
I manually built with an updated version and it compiles with Guava 27; tests running manually locally now.
Closes#26911 from srowen/SPARK-30272.
Authored-by: Sean Owen <srowen@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
1. Revert "Preparing development version 3.0.1-SNAPSHOT": 56dcd79
2. Revert "Preparing Spark release v3.0.0-preview2-rc2": c216ef1
### Why are the changes needed?
Shouldn't change master.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
manual test:
https://github.com/apache/spark/compare/5de5e46..wangyum:revert-masterCloses#26915 from wangyum/revert-master.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Yuming Wang <wgyumg@gmail.com>
### What changes were proposed in this pull request?
To push the built jars to maven release repository, we need to remove the 'SNAPSHOT' tag from the version name.
Made the following changes in this PR:
* Update all the `3.0.0-SNAPSHOT` version name to `3.0.0-preview`
* Update the sparkR version number check logic to allow jvm version like `3.0.0-preview`
**Please note those changes were generated by the release script in the past, but this time since we manually add tags on master branch, we need to manually apply those changes too.**
We shall revert the changes after 3.0.0-preview release passed.
### Why are the changes needed?
To make the maven release repository to accept the built jars.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
N/A
### What changes were proposed in this pull request?
To push the built jars to maven release repository, we need to remove the 'SNAPSHOT' tag from the version name.
Made the following changes in this PR:
* Update all the `3.0.0-SNAPSHOT` version name to `3.0.0-preview`
* Update the PySpark version from `3.0.0.dev0` to `3.0.0`
**Please note those changes were generated by the release script in the past, but this time since we manually add tags on master branch, we need to manually apply those changes too.**
We shall revert the changes after 3.0.0-preview release passed.
### Why are the changes needed?
To make the maven release repository to accept the built jars.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
N/A
Closes#26243 from jiangxb1987/3.0.0-preview-prepare.
Lead-authored-by: Xingbo Jiang <xingbo.jiang@databricks.com>
Co-authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: Xingbo Jiang <xingbo.jiang@databricks.com>
## What changes were proposed in this pull request?
After SPARK-27677, the shuffle client not only handles the shuffle block but also responsible for local persist RDD blocks. For better code scalability and precise semantics(as the [discussion](https://github.com/apache/spark/pull/24892#discussion_r300173331)), here we did several changes:
- Rename ShuffleClient to BlockStoreClient.
- Correspondingly rename the ExternalShuffleClient to ExternalBlockStoreClient, also change the server-side class from ExternalShuffleBlockHandler to ExternalBlockHandler.
- Move MesosExternalBlockStoreClient to Mesos package.
Note, we still keep the name of BlockTransferService, because the `Service` contains both client and server, also the name of BlockTransferService is not referencing shuffle client only.
## How was this patch tested?
Existing UT.
Closes#25327 from xuanyuanking/SPARK-28593.
Lead-authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Co-authored-by: Yuanjian Li <yuanjian.li@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
https://amplab.cs.berkeley.edu/jenkins/view/Spark%20QA%20Compile/job/spark-master-lint/
```
Checkstyle checks failed at following occurrences:
[ERROR] src/main/java/org/apache/spark/network/yarn/YarnShuffleServiceMetrics.java:[99] (sizes) LineLength: Line is longer than 100 characters (found 104).
[ERROR] src/main/java/org/apache/spark/network/yarn/YarnShuffleServiceMetrics.java:[101] (sizes) LineLength: Line is longer than 100 characters (found 101).
[ERROR] src/main/java/org/apache/spark/network/yarn/YarnShuffleServiceMetrics.java:[103] (sizes) LineLength: Line is longer than 100 characters (found 102).
[ERROR] src/main/java/org/apache/spark/network/yarn/YarnShuffleServiceMetrics.java:[105] (sizes) LineLength: Line is longer than 100 characters (found 103).
```
## How was this patch tested?
N/A
Closes#24760 from gatorsmile/updateYarnShuffleServiceMetrics.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Add a metric for number of exceptions caught in the `ExternalShuffleBlockHandler`, the idea being that spikes in this metric over some time window (or more desirably, the lack thereof) can be used as an indicator of the health of an external shuffle service. (Where "health" refers to its ability to successfully respond to client requests.)
## How was this patch tested?
Deployed a build of this PR to a YARN cluster, and confirmed that the NodeManagers' JMX metrics include `numCaughtExceptions`.
Closes#24645 from sjrand/SPARK-27773.
Authored-by: Steven Rand <srand@palantir.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
This pr removes duplicate declaration of plugin `org.apache.maven.plugins:maven-antrun-plugin`:
```
[WARNING] Some problems were encountered while building the effective model for org.apache.spark:spark-network-yarn_2.12🫙3.0.0-SNAPSHOT
[WARNING] 'build.plugins.plugin.(groupId:artifactId)' must be unique but found duplicate declaration of plugin org.apache.maven.plugins:maven-antrun-plugin line 177, column 15
[WARNING]
[WARNING] It is highly recommended to fix these problems because they threaten the stability of your build.
[WARNING]
[WARNING] For this reason, future Maven versions might no longer support building such malformed projects.
[WARNING]
```
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/105523/consoleFull
## How was this patch tested?
Existing test
Closes#24641 from wangyum/SPARK-27610.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Fixed the `spark-<version>-yarn-shuffle.jar` artifact packaging to shade the native netty libraries:
- shade the `META-INF/native/libnetty_*` native libraries when packagin
the yarn shuffle service jar. This is required as netty library loader
derives that based on shaded package name.
- updated the `org/spark_project` shade package prefix to `org/sparkproject`
(i.e. removed underscore) as the former breaks the netty native lib loading.
This was causing the yarn external shuffle service to fail
when spark.shuffle.io.mode=EPOLL
## How was this patch tested?
Manual tests
Closes#24502 from amuraru/SPARK-27610_master.
Authored-by: Adi Muraru <amuraru@adobe.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
Creating an Netty `EventLoopGroup` leads to creating a new Thread pool for handling the events. For stopping the threads of the pool the event loop group should be shut down which is properly done for transport servers and clients by calling for example the `shutdownGracefully()` method (for details see the `close()` method of `TransportClientFactory` and `TransportServer`). But there is a separate event loop group for shuffle chunk fetch requests which is in pipeline for handling fetch request (shared between the client and server) and owned by the `TransportContext` and this was never shut down.
## How was this patch tested?
With existing unittest.
This leak is in the production system too but its effect is spiking in the unittest.
Checking the core unittest logs before the PR:
```
$ grep "LEAK IN SUITE" unit-tests.log | grep -o shuffle-chunk-fetch-handler | wc -l
381
```
And after the PR without whitelisting in thread audit and with an extra `await` after the
` chunkFetchWorkers.shutdownGracefully()`:
```
$ grep "LEAK IN SUITE" unit-tests.log | grep -o shuffle-chunk-fetch-handler | wc -l
0
```
Closes#23930 from attilapiros/SPARK-27021.
Authored-by: “attilapiros” <piros.attila.zsolt@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
Recently, the ability to expose the metrics for YARN Shuffle Service was added as part of [SPARK-18364](https://github.com/apache/spark/pull/22485). We need to add some metrics to be able to determine the number of active connections as well as open connections to the external shuffle service to benchmark network and connection issues on large cluster environments.
Added two more shuffle server metrics for Spark Yarn shuffle service: numRegisteredConnections which indicate the number of registered connections to the shuffle service and numActiveConnections which indicate the number of active connections to the shuffle service at any given point in time.
If these metrics are outputted to a file, we get something like this:
1533674653489 default.shuffleService: Hostname=server1.abc.com, openBlockRequestLatencyMillis_count=729, openBlockRequestLatencyMillis_rate15=0.7110833548897356, openBlockRequestLatencyMillis_rate5=1.657808981793011, openBlockRequestLatencyMillis_rate1=2.2404486061620474, openBlockRequestLatencyMillis_rateMean=0.9242558551196706,
numRegisteredConnections=35,
blockTransferRateBytes_count=2635880512, blockTransferRateBytes_rate15=2578547.6094160094, blockTransferRateBytes_rate5=6048721.726302424, blockTransferRateBytes_rate1=8548922.518223226, blockTransferRateBytes_rateMean=3341878.633637769, registeredExecutorsSize=5, registerExecutorRequestLatencyMillis_count=5, registerExecutorRequestLatencyMillis_rate15=0.0027973949328659836, registerExecutorRequestLatencyMillis_rate5=0.0021278007987206426, registerExecutorRequestLatencyMillis_rate1=2.8270296777387467E-6, registerExecutorRequestLatencyMillis_rateMean=0.006339206380043053, numActiveConnections=35
Closes#22498 from pgandhi999/SPARK-18364.
Authored-by: pgandhi <pgandhi@oath.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
This PR makes Spark's default Scala version as 2.12, and Scala 2.11 will be the alternative version. This implies that Scala 2.12 will be used by our CI builds including pull request builds.
We'll update the Jenkins to include a new compile-only jobs for Scala 2.11 to ensure the code can be still compiled with Scala 2.11.
## How was this patch tested?
existing tests
Closes#22967 from dbtsai/scala2.12.
Authored-by: DB Tsai <d_tsai@apple.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
This patch is to bump the master branch version to 3.0.0-SNAPSHOT.
## How was this patch tested?
N/A
Closes#22606 from gatorsmile/bump3.0.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
This PR is follow-up of closed https://github.com/apache/spark/pull/17401 which only ended due to of inactivity, but its still nice feature to have.
Given review by jerryshao taken in consideration and edited:
- VisibleForTesting deleted because of dependency conflicts
- removed unnecessary reflection for `MetricsSystemImpl`
- added more available types for gauge
## How was this patch tested?
Manual deploy of new yarn-shuffle jar into a Node Manager and verifying that the metrics appear in the Node Manager-standard location. This is JMX with an query endpoint running on `hostname:port`
Resulting metrics look like this:
```
curl -sk -XGET hostname:port | grep -v '#' | grep 'shuffleService'
hadoop_nodemanager_openblockrequestlatencymillis_rate15{name="shuffleService",} 0.31428910657834713
hadoop_nodemanager_blocktransferratebytes_rate15{name="shuffleService",} 566144.9983653595
hadoop_nodemanager_blocktransferratebytes_ratemean{name="shuffleService",} 2464409.9678099006
hadoop_nodemanager_openblockrequestlatencymillis_rate1{name="shuffleService",} 1.2893844732240272
hadoop_nodemanager_registeredexecutorssize{name="shuffleService",} 2.0
hadoop_nodemanager_openblockrequestlatencymillis_ratemean{name="shuffleService",} 1.255574678369966
hadoop_nodemanager_openblockrequestlatencymillis_count{name="shuffleService",} 315.0
hadoop_nodemanager_openblockrequestlatencymillis_rate5{name="shuffleService",} 0.7661929192569739
hadoop_nodemanager_registerexecutorrequestlatencymillis_ratemean{name="shuffleService",} 0.0
hadoop_nodemanager_registerexecutorrequestlatencymillis_count{name="shuffleService",} 0.0
hadoop_nodemanager_registerexecutorrequestlatencymillis_rate1{name="shuffleService",} 0.0
hadoop_nodemanager_registerexecutorrequestlatencymillis_rate5{name="shuffleService",} 0.0
hadoop_nodemanager_blocktransferratebytes_count{name="shuffleService",} 6.18271213E8
hadoop_nodemanager_registerexecutorrequestlatencymillis_rate15{name="shuffleService",} 0.0
hadoop_nodemanager_blocktransferratebytes_rate5{name="shuffleService",} 1154114.4881816586
hadoop_nodemanager_blocktransferratebytes_rate1{name="shuffleService",} 574745.0749848988
```
Closes#22485 from mareksimunek/SPARK-18364.
Lead-authored-by: marek.simunek <marek.simunek@firma.seznam.cz>
Co-authored-by: Andrew Ash <andrew@andrewash.com>
Signed-off-by: Thomas Graves <tgraves@apache.org>
## What changes were proposed in this pull request?
In the dev list, we can still discuss whether the next version is 2.5.0 or 3.0.0. Let us first bump the master branch version to `2.5.0-SNAPSHOT`.
## How was this patch tested?
N/A
Closes#22426 from gatorsmile/bumpVersionMaster.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
This patch bumps the master branch version to `2.4.0-SNAPSHOT`.
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20222 from gatorsmile/bump24.
In the current code, if NM recovery is not enabled then `YarnShuffleService` will write shuffle metadata to NM local dir-1, if this local dir-1 is on bad disk, then `YarnShuffleService` will be failed to start. So to solve this issue, in Spark side if NM recovery is not enabled, then Spark will not persist data into leveldb, in that case yarn shuffle service can still be served but lose the ability for recovery, (it is fine because the failure of NM will kill the containers as well as applications).
Tested in the local cluster with NM recovery off and on to see if folder is created or not. MiniCluster UT isn't added because in MiniCluster NM will always set port to 0, but NM recovery requires non-ephemeral port.
Author: jerryshao <sshao@hortonworks.com>
Closes#19032 from jerryshao/SPARK-17321.
Change-Id: I8f2fe73d175e2ad2c4e380caede3873e0192d027
There was some code based on the old SASL handler in the new auth client that
was incorrectly using the SASL user as the user to authenticate against the
external shuffle service. This caused the external service to not be able to
find the correct secret to authenticate the connection, failing the connection.
In the course of debugging, I found that some log messages from the YARN shuffle
service were a little noisy, so I silenced some of them, and also added a couple
of new ones that helped find this issue. On top of that, I found that a check
in the code that records app secrets was wrong, causing more log spam and also
using an O(n) operation instead of an O(1) call.
Also added a new integration suite for the YARN shuffle service with auth on,
and verified it failed before, and passes now.
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#18706 from vanzin/SPARK-21494.
## What changes were proposed in this pull request?
Remove all usages of Scala Tuple2 from common/network-* projects. Otherwise, Yarn users cannot use `spark.reducer.maxReqSizeShuffleToMem`.
## How was this patch tested?
Jenkins.
Author: Shixiong Zhu <shixiong@databricks.com>
Closes#18593 from zsxwing/SPARK-21369.
and contains scala classes
## What changes were proposed in this pull request?
This change ensures that all references to guava from within the yarn shuffle jar pointed to the shaded guava class already provided in the jar.
Also, it explicitly excludes scala classes from being added to the jar.
## How was this patch tested?
Ran unit tests on the module and they passed.
javap now returns the expected result - reference to the shaded guava under `org/spark_project` (previously this was referring to `com.google...`
```
javap -cp common/network-yarn/target/scala-2.11/spark-2.3.0-SNAPSHOT-yarn-shuffle.jar -c org/apache/spark/network/yarn/YarnShuffleService | grep Lists
57: invokestatic #138 // Method org/spark_project/guava/collect/Lists.newArrayList:()Ljava/util/ArrayList;
```
Guava is still shaded in the jar:
```
jar -tf common/network-yarn/target/scala-2.11/spark-2.3.0-SNAPSHOT-yarn-shuffle.jar | grep guava | head
META-INF/maven/com.google.guava/
META-INF/maven/com.google.guava/guava/
META-INF/maven/com.google.guava/guava/pom.properties
META-INF/maven/com.google.guava/guava/pom.xml
org/spark_project/guava/
org/spark_project/guava/annotations/
org/spark_project/guava/annotations/Beta.class
org/spark_project/guava/annotations/GwtCompatible.class
org/spark_project/guava/annotations/GwtIncompatible.class
org/spark_project/guava/annotations/VisibleForTesting.class
```
(not sure if the above META-INF/* is a problem or not)
I took this jar, deployed it on a yarn cluster with shuffle service enabled, and made sure the YARN node managers came up. An application with a shuffle was run and it succeeded.
Author: Mark Grover <mark@apache.org>
Closes#17990 from markgrover/spark-20756.
## What changes were proposed in this pull request?
Fix build warnings primarily related to Breeze 0.13 operator changes, Java style problems
## How was this patch tested?
Existing tests
Author: Sean Owen <sowen@cloudera.com>
Closes#17803 from srowen/SPARK-20523.
…ss NFS directories
## What changes were proposed in this pull request?
Change from using java Files.move to use Hadoop filesystem operations to move the directories. The java Files.move does not work when moving directories across NFS mounts and in fact also says that if the directory has entries you should do a recursive move. We are already using Hadoop filesystem here so just use the local filesystem from there as it handles this properly.
Note that the DB here is actually a directory of files and not just a single file, hence the change in the name of the local var.
## How was this patch tested?
Ran YarnShuffleServiceSuite unit tests. Unfortunately couldn't easily add one here since involves NFS.
Ran manual tests to verify that the DB directories were properly moved across NFS mounted directories. Have been running this internally for weeks.
Author: Tom Graves <tgraves@apache.org>
Closes#17748 from tgravescs/SPARK-19812.
## What changes were proposed in this pull request?
Avoid `NoSuchElementException` every time `ConfigProvider.get(val, default)` falls back to default. This apparently causes non-trivial overhead in at least one path, and can easily be avoided.
See https://github.com/apache/spark/pull/17329
## How was this patch tested?
Existing tests
Author: Sean Owen <sowen@cloudera.com>
Closes#17567 from srowen/SPARK-19991.
This change introduces a new auth mechanism to the transport library,
to be used when users enable strong encryption. This auth mechanism
has better security than the currently used DIGEST-MD5.
The new protocol uses symmetric key encryption to mutually authenticate
the endpoints, and is very loosely based on ISO/IEC 9798.
The new protocol falls back to SASL when it thinks the remote end is old.
Because SASL does not support asking the server for multiple auth protocols,
which would mean we could re-use the existing SASL code by just adding a
new SASL provider, the protocol is implemented outside of the SASL API
to avoid the boilerplate of adding a new provider.
Details of the auth protocol are discussed in the included README.md
file.
This change partly undos the changes added in SPARK-13331; AES encryption
is now decoupled from SASL authentication. The encryption code itself,
though, has been re-used as part of this change.
## How was this patch tested?
- Unit tests
- Tested Spark 2.2 against Spark 1.6 shuffle service with SASL enabled
- Tested Spark 2.2 against Spark 2.2 shuffle service with SASL fallback disabled
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#16521 from vanzin/SPARK-19139.
Remove spark-tag's compile-scope dependency (and, indirectly, spark-core's compile-scope transitive-dependency) on scalatest by splitting test-oriented tags into spark-tags' test JAR.
Alternative to #16303.
Author: Ryan Williams <ryan.blake.williams@gmail.com>
Closes#16311 from ryan-williams/tt.
This change moves the logic that translates Spark configuration to
commons-crypto configuration to the network-common module. It also
extends TransportConf and ConfigProvider to provide the necessary
interfaces for the translation to work.
As part of the change, I removed SystemPropertyConfigProvider, which
was mostly used as an "empty config" in unit tests, and adjusted the
very few tests that required a specific config.
I also changed the config keys for AES encryption to live under the
"spark.network." namespace, which is more correct than their previous
names under "spark.authenticate.".
Tested via existing unit test.
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#16200 from vanzin/SPARK-18773.
## What changes were proposed in this pull request?
This patch bumps master branch version to 2.2.0-SNAPSHOT.
## How was this patch tested?
N/A
Author: Reynold Xin <rxin@databricks.com>
Closes#16126 from rxin/SPARK-18695.
Currently, the code is just swallowing exceptions, and not really checking
whether the auth information was being recorded properly. Fix both problems,
and also avoid tests inadvertently affecting other tests by modifying the
shared config variable (by making it not shared).
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#15161 from vanzin/SPARK-17611.
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.
The Spark Yarn Shuffle Service doesn't re-initialize the application credentials early enough which causes any other spark executors trying to fetch from that node during a rolling upgrade to fail with "java.lang.NullPointerException: Password cannot be null if SASL is enabled". Right now the spark shuffle service relies on the Yarn nodemanager to re-register the applications, unfortunately this is after we open the port for other executors to connect. If other executors connected before the re-register they get a null pointer exception which isn't a re-tryable exception and cause them to fail pretty quickly. To solve this I added another leveldb file so that it can save and re-initialize all the applications before opening the port for other executors to connect to it. Adding another leveldb was simpler from the code structure point of view.
Most of the code changes are moving things to common util class.
Patch was tested manually on a Yarn cluster with rolling upgrade was happing while spark job was running. Without the patch I consistently get the NullPointerException, with the patch the job gets a few Connection refused exceptions but the retries kick in and the it succeeds.
Author: Thomas Graves <tgraves@staydecay.corp.gq1.yahoo.com>
Closes#14718 from tgravescs/SPARK-16711.
## What changes were proposed in this pull request?
Make all Java Loggers static members
## How was this patch tested?
Jenkins
Author: Sean Owen <sowen@cloudera.com>
Closes#14896 from srowen/SPARK-17332.
https://issues.apache.org/jira/browse/SPARK-16535
## What changes were proposed in this pull request?
When I scan through the pom.xml of sub projects, I found this warning as below and attached screenshot
```
Definition of groupId is redundant, because it's inherited from the parent
```
I've tried to remove some of the lines with groupId definition, and the build on my local machine is still ok.
```
<groupId>org.apache.spark</groupId>
```
As I just find now `<maven.version>3.3.9</maven.version>` is being used in Spark 2.x, and Maven-3 supports versionless parent elements: Maven 3 will remove the need to specify the parent version in sub modules. THIS is great (in Maven 3.1).
ref: http://stackoverflow.com/questions/3157240/maven-3-worth-it/3166762#3166762
## How was this patch tested?
I've tested by re-building the project, and build succeeded.
Author: Xin Ren <iamshrek@126.com>
Closes#14189 from keypointt/SPARK-16535.
