You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by uncleGen <gi...@git.apache.org> on 2014/07/10 14:43:41 UTC
[GitHub] spark pull request: Bug Fix: LiveListenerBus Queue Overflow
GitHub user uncleGen opened a pull request:
https://github.com/apache/spark/pull/1356
Bug Fix: LiveListenerBus Queue Overflow
As we know, the size of "eventQueue" is fixed. When event comes faster than consume speed of listener, overflow events will be thrown away with throwing an "logQueueFullErrorMessage'. As a result, we find spark UI in chaotic state. In my Strategy, those overflow events are "cached" in "blockManager" first , and replayed by "replayThread" when "eventQueue" is free. The unique drawback is some display delay in Spark UI. But it is better than event loss. This delay will disappear when event stress alleviates.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/uncleGen/spark master
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/1356.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 #1356
----
commit 52906a911b111f3848e21c3b449b76237f658051
Author: uncleGen <hu...@gmail.com>
Date: 2014-07-10T12:17:32Z
Bug Fix: LiveListenerBus Queue Overflow
----
---
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.
---
[GitHub] spark pull request: [SPARK-2675]LiveListenerBus Queue Overflow
Posted by pwendell <gi...@git.apache.org>.
Github user pwendell commented on the pull request:
https://github.com/apache/spark/pull/1356#issuecomment-53977396
Okay I think this is no longer necessary now that we fixed the issue causing lag in processing events. So I'd like to close this issue for now.
---
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: Bug Fix: LiveListenerBus Queue Overflow
Posted by uncleGen <gi...@git.apache.org>.
Github user uncleGen commented on the pull request:
https://github.com/apache/spark/pull/1356#issuecomment-48691344
@pwendell yeah, this is not a handsome way to resolve the bug. My fix is a compromised way. Actually, there are no frequent get/put opertions in blockManager when "caching" overflow events. In my test, it can alleviates event stress. As you said, improving consume speed of listener is a better way.　But, I have no confidence whether your way can adapt the scene where 100000+ events comes instantaneously.
---
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.
---
[GitHub] spark pull request: Bug Fix: LiveListenerBus Queue Overflow
Posted by pwendell <gi...@git.apache.org>.
Github user pwendell commented on the pull request:
https://github.com/apache/spark/pull/1356#issuecomment-48634350
The underlying issue here is that one of the listener implementations is inefficient and can cause jobs to ended up in an unstable queue: https://issues.apache.org/jira/browse/SPARK-2316
This is alleviated but not fully fixed in the 1.0.1 release.
Putting the events in the blockmanager is very expensive, so I don't think it's a good solution here. We should just fix the listeners so that they are fast enough to keep up.
---
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.
---
[GitHub] spark pull request: Bug Fix: LiveListenerBus Queue Overflow
Posted by uncleGen <gi...@git.apache.org>.
Github user uncleGen commented on the pull request:
https://github.com/apache/spark/pull/1356#issuecomment-48691872
@pwendell yeah, it is not a handsome way to resolve the bug. My fix is a compromise way. Actualy, is will not cause frequent put/get opertions in blockManager when "caching" overflow events. In my test, it can alleviate event stress effectively. As you said, improving consume speed of listener is a better way, and i agree. But, i have no confidence whether it can adapt the scene where 100000+ events come instantaneously
---
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.
---
[GitHub] spark pull request: Bug Fix: LiveListenerBus Queue Overflow
Posted by AmplabJenkins <gi...@git.apache.org>.
Github user AmplabJenkins commented on the pull request:
https://github.com/apache/spark/pull/1356#issuecomment-48599352
Can one of the admins verify this patch?
---
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.
---
[GitHub] spark pull request: [SPARK-2675]LiveListenerBus Queue Overflow
Posted by asfgit <gi...@git.apache.org>.
Github user asfgit closed the pull request at:
https://github.com/apache/spark/pull/1356
---
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-2675]LiveListenerBus Queue Overflow
Posted by uncleGen <gi...@git.apache.org>.
Github user uncleGen commented on the pull request:
https://github.com/apache/spark/pull/1356#issuecomment-53977557
okay!
---
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