You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by lianhuiwang <gi...@git.apache.org> on 2015/02/09 16:41:27 UTC
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter need to cat...
GitHub user lianhuiwang opened a pull request:
https://github.com/apache/spark/pull/4474
[SPARK-5687][Core]TaskResultGetter need to catch OutOfMemoryError.
because in enqueueSuccessfulTask there is another thread to fetch result, if result is large,it maybe throw a OutOfMemoryError. so if we donot catch OutOfMemoryError, DAGDAGScheduler donot know the status of this task.
@andrewor14
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/lianhuiwang/spark SPARK-5687
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/4474.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #4474
----
commit 7d538fc4155b1425613390e3b5d1db990b9b1ef2
Author: lianhuiwang <li...@gmail.com>
Date: 2015-02-09T15:38:34Z
fix OutOfMemoryError
----
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter needs to ca...
Posted by lianhuiwang <gi...@git.apache.org>.
Github user lianhuiwang commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-95907922
@pwendell i think we cannot kill JVM directly when this occurs. when it is hive server that one driver for many jobs, if we kill JVM, other jobs on this driver cannot continue. i think this pr is ok that just abort this job and then DAGScheduler will throw a jobFailed exception to client. if it is hive server, then hive server can catch this exception and continue to run other jobs. if it is a application that i said, user application donot catch this exception and throw this to applicationMaster, then application will be failed. so that can ensure that is be right for any situations.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter needs to ca...
Posted by lianhuiwang <gi...@git.apache.org>.
Github user lianhuiwang commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-95773327
@pwendell @andrewor14 i think i should reopen this PR. because i got this error yesterday. when i used collect() from many executors. then task-result-getter thread catch the exception "java.lang.OutOfMemoryError: Java heap space". but driver has been running. now my application run on yarn-cluster. so application client think that application is running but actually it has error and client donot know it. so i think driver/application master should abort its tasks and let client know that driver has errors.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter needs to ca...
Posted by lianhuiwang <gi...@git.apache.org>.
Github user lianhuiwang closed the pull request at:
https://github.com/apache/spark/pull/4474
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter needs to ca...
Posted by srowen <gi...@git.apache.org>.
Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73817485
If this is a WontFix, do you mind closing this PR?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter need to cat...
Posted by SparkQA <gi...@git.apache.org>.
Github user SparkQA commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73545420
[Test build #27106 has finished](https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/27106/consoleFull) for PR 4474 at commit [`7d538fc`](https://github.com/apache/spark/commit/7d538fc4155b1425613390e3b5d1db990b9b1ef2).
* This patch **passes all tests**.
* This patch merges cleanly.
* This patch adds no public classes.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter needs to ca...
Posted by pwendell <gi...@git.apache.org>.
Github user pwendell commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73632020
Yes, if you find a situation where the driver can oom, just let us know, we can try to fix it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter needs to ca...
Posted by pwendell <gi...@git.apache.org>.
Github user pwendell commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-95805067
I'd be happy to have a patch that kills the JVM when this occurs, with a warning message logged. I didn't realize in your original submission that this was actually just killing the thread but allowing the JVM to survive.
Really, once we are out of memory, we should coerce the JVM to terminate. I agree that is better than having a silent thread death.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter needs to ca...
Posted by lianhuiwang <gi...@git.apache.org>.
Github user lianhuiwang commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73627335
@andrewor14 @pwendell yes, Now we have conf.get("spark.driver.maxResultSize", "1g") to control the memory of driver. but once OOM is happened at TaskResultGetter, the application will be hold and waste resouces. so i think this is a very serious and at least we should to let application abort and let user know why his job is wrong.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter need to cat...
Posted by pwendell <gi...@git.apache.org>.
Github user pwendell commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73599694
Yes - the driver should terminate when there is an OOM, it is an uncrecoverable failure.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter need to cat...
Posted by SparkQA <gi...@git.apache.org>.
Github user SparkQA commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73530090
[Test build #27106 has started](https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/27106/consoleFull) for PR 4474 at commit [`7d538fc`](https://github.com/apache/spark/commit/7d538fc4155b1425613390e3b5d1db990b9b1ef2).
* This patch merges cleanly.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter need to cat...
Posted by andrewor14 <gi...@git.apache.org>.
Github user andrewor14 commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73564181
@lianhuiwang I think it's a bad idea in general to catch out of memory errors, because by the time it is thrown the JVM may already be in a bad state even if we are able to recover by aborting the task. I think the correct thing to do here is to avoid the OOM in the first place through some other means instead of trying to recover from it when it's already too late.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request: [SPARK-5687][Core]TaskResultGetter need to cat...
Posted by AmplabJenkins <gi...@git.apache.org>.
Github user AmplabJenkins commented on the pull request:
https://github.com/apache/spark/pull/4474#issuecomment-73545439
Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/27106/
Test PASSed.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org