Build #145673

Environment variables

NameValue
ANDROID_HOME/home/android-sdk/
AWS_ACCESS_KEY_ID[*******]
AWS_SECRET_ACCESS_KEY[*******]
BUILD_CAUSEGHPRBCAUSE
BUILD_CAUSE_GHPRBCAUSEtrue
BUILD_DISPLAY_NAME#145673
BUILD_ID145673
BUILD_NUMBER145673
BUILD_TAGjenkins-SparkPullRequestBuilder-145673
BUILD_URLhttps://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/145673/
CLASSPATH$CLASSPATH
DBUS_SESSION_BUS_ADDRESSunix:path=/run/user/1001/bus
EXECUTOR_NUMBER1
GITHUB_OAUTH_KEY[*******]
GIT_BRANCHSPARK-33701
GIT_COMMITb6340e82d6910e43b8cde95c3036c7c0252e0d74
GIT_PREVIOUS_COMMITa46db3fb2ead423754d1d5add48cbdc9862cc77b
GIT_PREVIOUS_SUCCESSFUL_COMMITa46db3fb2ead423754d1d5add48cbdc9862cc77b
GIT_URLhttps://github.com/apache/spark.git
HOME/home/jenkins
HUDSON_HOME/var/lib/jenkins
HUDSON_SERVER_COOKIE472906e9832aeb79
HUDSON_URLhttps://amplab.cs.berkeley.edu/jenkins/
JAVA_HOME/usr/java/latest
JENKINS_HOME/var/lib/jenkins
JENKINS_SERVER_COOKIE472906e9832aeb79
JENKINS_URLhttps://amplab.cs.berkeley.edu/jenkins/
JOB_BASE_NAMESparkPullRequestBuilder
JOB_NAMESparkPullRequestBuilder
JOB_URLhttps://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/
LANGen_US.UTF-8
LOGNAMEjenkins
MOTD_SHOWNpam
NODE_LABELSresearch-jenkins-worker-08 ubuntu ubuntu20 worker-08
NODE_NAMEresearch-jenkins-worker-08
OLDPWD/home/jenkins
PATH/home/jenkins/tools/hudson.tasks.Maven_MavenInstallation/Maven_3.6.3/bin/:/home/jenkins/gems/bin:/usr/local/go/bin:/home/jenkins/go-projects/bin:/home/jenkins/anaconda2/bin:/home/jenkins/tools/hudson.tasks.Maven_MavenInstallation/Maven_3.6.3/bin/:/home/jenkins/gems/bin:/usr/local/go/bin:/home/jenkins/go-projects/bin:/home/jenkins/anaconda2/bin:$PATH
PWD/home/jenkins
ROOT_BUILD_CAUSEGHPRBCAUSE
ROOT_BUILD_CAUSE_GHPRBCAUSEtrue
RUN_ARTIFACTS_DISPLAY_URLhttps://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/145673/display/redirect?page=artifacts
RUN_CHANGES_DISPLAY_URLhttps://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/145673/display/redirect?page=changes
RUN_DISPLAY_URLhttps://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/145673/display/redirect
RUN_TESTS_DISPLAY_URLhttps://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/145673/display/redirect?page=tests
SHELL/bin/bash
SHLVL0
SPARK_TEST_KEY[*******]
SSH_CLIENT192.168.10.11 35144 22
SSH_CONNECTION192.168.10.11 35144 192.168.10.28 22
USERjenkins
WORKSPACE/home/jenkins/workspace/SparkPullRequestBuilder
XDG_RUNTIME_DIR/run/user/1001
XDG_SESSION_CLASSuser
XDG_SESSION_ID1
XDG_SESSION_TYPEtty
_/usr/java/latest/bin/java
ghprbActualCommit93d2470c994f89dbbe6258ee7aaa254144a9bb35
ghprbActualCommitAuthorVenkata krishnan Sowrirajan
ghprbActualCommitAuthorEmailvsowrirajan@linkedin.com
ghprbAuthorRepoGitUrlhttps://github.com/venkata91/spark.git
ghprbCommentBodynull
ghprbCredentialsIdb7d94526-9e9b-435f-9275-d7dbf209f4a3
ghprbGhRepositoryapache/spark
ghprbPullAuthorEmailvsowrirajan@linkedin.com
ghprbPullAuthorLoginvenkata91
ghprbPullAuthorLoginMention@venkata91
ghprbPullDescriptionGitHub pull request #33896 of commit 93d2470c994f89dbbe6258ee7aaa254144a9bb35, no merge conflicts.
ghprbPullId33896
ghprbPullLinkhttps://github.com/apache/spark/pull/33896
ghprbPullLongDescription<!--\r\nThanks for sending a pull request!  Here are some tips for you:\r\n  1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html\r\n  2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html\r\n  3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'.\r\n  4. Be sure to keep the PR description updated to reflect all changes.\r\n  5. Please write your PR title to summarize what this PR proposes.\r\n  6. If possible, provide a concise example to reproduce the issue for a faster review.\r\n  7. If you want to add a new configuration, please read the guideline first for naming configurations in\r\n     'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.\r\n  8. If you want to add or modify an error type or message, please read the guideline first in\r\n     'core/src/main/resources/error/README.md'.\r\n-->\r\n\r\n### What changes were proposed in this pull request?\r\n<!--\r\nPlease clarify what changes you are proposing. The purpose of this section is to outline the changes and how this PR fixes the issue. \r\nIf possible, please consider writing useful notes for better and faster reviews in your PR. See the examples below.\r\n  1. If you refactor some codes with changing classes, showing the class hierarchy will help reviewers.\r\n  2. If you fix some SQL features, you can provide some references of other DBMSes.\r\n  3. If there is design documentation, please add the link.\r\n  4. If there is a discussion in the mailing list, please add the link.\r\n-->\r\nAs part of SPARK-32920 implemented a simple approach to finalization for push-based shuffle. Shuffle merge finalization is the final operation happens at the end of the stage when all the tasks are completed asking all the external shuffle services to complete the shuffle merge for the stage. Once this request is completed no more shuffle pushes will be accepted. With this approach, `DAGScheduler` waits for a fixed time of 10s (`spark.shuffle.push.finalize.timeout`) to allow some time for the inflight shuffle pushes to complete, but this adds additional overhead to stages with very little shuffles.\r\n\r\nIn this PR, instead of waiting for fixed amount of time before shuffle merge finalization now this is controlled adaptively if min threshold number of map tasks shuffle push (`spark.shuffle.push.minPushRatio`) completed then shuffle merge finalization will be scheduled. Also additionally if the total shuffle generated is lesser than min threshold shuffle size (`spark.shuffle.push.minShuffleSizeToWait`) then immediately shuffle merge finalization is scheduled.\r\n### Why are the changes needed?\r\n<!--\r\nPlease clarify why the changes are needed. For instance,\r\n  1. If you propose a new API, clarify the use case for a new API.\r\n  2. If you fix a bug, you can clarify why it is a bug.\r\n-->\r\nThis is a performance improvement to the existing functionality\r\n\r\n### Does this PR introduce _any_ user-facing change?\r\n<!--\r\nNote that it means *any* user-facing change including all aspects such as the documentation fix.\r\nIf yes, please clarify the previous behavior and the change this PR proposes - provide the console output, description and/or an example to show the behavior difference if possible.\r\nIf possible, please also clarify if this is a user-facing change compared to the released Spark versions or within the unreleased branches such as master.\r\nIf no, write 'No'.\r\n-->\r\nYes additional user facing configs `spark.shuffle.push.minPushRatio` and `spark.shuffle.push.minShuffleSizeToWait`\r\n\r\n### How was this patch tested?\r\n<!--\r\nIf tests were added, say they were added here. Please make sure to add some test cases that check the changes thoroughly including negative and positive cases if possible.\r\nIf it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future.\r\nIf tests were not added, please describe why they were not added and/or why it was difficult to add.\r\n-->\r\nAdded unit tests in `DAGSchedulerSuite`\r\n\r\nLead-authored-by: Min Shen <mshen@linkedin.com>\r\nCo-authored-by: Venkata krishnan Sowrirajan <vsowrirajan@linkedin.com>
ghprbPullTitle[SPARK-33701][SHUFFLE] Adaptive shuffle merge finalization for push-based shuffle
ghprbSourceBranchSPARK-33701
ghprbTargetBranchmaster
ghprbTriggerAuthor
ghprbTriggerAuthorEmail
ghprbTriggerAuthorLogin
ghprbTriggerAuthorLoginMention
sha1origin/pr/33896/merge