Fixed bug for Jira Issue# 1527 #517
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Details: rootDirs in DiskBlockManagerSuite doesn't get full path from rootDir0, rootDir1
|
I have emailed the person who has opened this Jira ...I am waiting for him to assign it to me...if someone can assign it to me thn I can change Jira bug status appropritely. Email: [email protected] Thank you, |
|
Can one of the admins verify this patch? |
|
Jenkins, test this please. |
|
Merged build triggered. |
|
Merged build started. |
|
Merged build finished. |
|
Refer to this link for build results: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/14426/ |
|
Jenkins, retest this please. |
|
Merged build triggered. |
|
Merged build started. |
|
Merged build finished. |
|
Refer to this link for build results: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/14427/ |
There was a problem hiding this comment.
I don't think this import is required (and it's causing a compile error anyway).
|
removed import statement causing compilation error and build it locally...seems to be fixed now. |
|
build passed. |
pwendell
pushed a commit
to pwendell/spark
that referenced
this pull request
Apr 27, 2014
Move the PR#517 of apache-incubator-spark to the apache-spark Author: Mridul Muralidharan <[email protected]> Closes apache#159 from mridulm/master and squashes the following commits: 5ff59c2 [Mridul Muralidharan] Change property in suite also 167fad8 [Mridul Muralidharan] Address review comments 9bda70e [Mridul Muralidharan] Address review comments, akwats add to failedExecutors 270d841 [Mridul Muralidharan] Address review comments fa5d9f1 [Mridul Muralidharan] Bugfixes/improvements to scheduler : PR apache#517
yifeih
pushed a commit
to yifeih/spark
that referenced
this pull request
May 8, 2019
Implements the shuffle locations API as part of SPARK-25299. This adds an additional field to all `MapStatus` objects: a `MapShuffleLocations` that indicates where a task's map output is stored. This module is optional and implementations of the pluggable shuffle writers and readers can ignore it accordingly. This API is designed with the use case in mind of future plugin implementations desiring to have the driver store metadata about where shuffle blocks are stored. There are a few caveats to this design: - We originally wanted to remove the `BlockManagerId` from `MapStatus` entirely and replace it with this object. However, doing this proves to be very difficult, as many places use the block manager ID for other kinds of shuffle data bookkeeping. As a result, we concede to storing the block manager ID redundantly here. However, the overhead should be minimal: because we cache block manager ids and default map shuffle locations, the two fields in `MapStatus` should point to the same object on the heap. Thus we add `O(M)` storage overhead on the driver, where for each map status we're storing an additional pointer to the same on-heap object. We will run benchmarks against the TPC-DS workload to see if there are significant performance repercussions for this implementation. - `KryoSerializer` expects `CompressedMapStatus` and `HighlyCompressedMapStatus` to be serialized via reflection, so originally all fields of these classes needed to be registered with Kryo. However, the `MapShuffleLocations` is now pluggable. We think however that previously Kryo was defaulting to Java serialization anyways, so we now just explicitly tell Kryo to use `ExternalizableSerializer` to deal with these objects. There's a small hack in the serialization protocol that attempts to avoid serializing the same `BlockManagerId` twice in the case that the map shuffle locations is a `DefaultMapShuffleLocations`.
squito
pushed a commit
to squito/spark
that referenced
this pull request
Jul 18, 2019
Implements the shuffle locations API as part of SPARK-25299. This adds an additional field to all `MapStatus` objects: a `MapShuffleLocations` that indicates where a task's map output is stored. This module is optional and implementations of the pluggable shuffle writers and readers can ignore it accordingly. This API is designed with the use case in mind of future plugin implementations desiring to have the driver store metadata about where shuffle blocks are stored. There are a few caveats to this design: - We originally wanted to remove the `BlockManagerId` from `MapStatus` entirely and replace it with this object. However, doing this proves to be very difficult, as many places use the block manager ID for other kinds of shuffle data bookkeeping. As a result, we concede to storing the block manager ID redundantly here. However, the overhead should be minimal: because we cache block manager ids and default map shuffle locations, the two fields in `MapStatus` should point to the same object on the heap. Thus we add `O(M)` storage overhead on the driver, where for each map status we're storing an additional pointer to the same on-heap object. We will run benchmarks against the TPC-DS workload to see if there are significant performance repercussions for this implementation. - `KryoSerializer` expects `CompressedMapStatus` and `HighlyCompressedMapStatus` to be serialized via reflection, so originally all fields of these classes needed to be registered with Kryo. However, the `MapShuffleLocations` is now pluggable. We think however that previously Kryo was defaulting to Java serialization anyways, so we now just explicitly tell Kryo to use `ExternalizableSerializer` to deal with these objects. There's a small hack in the serialization protocol that attempts to avoid serializing the same `BlockManagerId` twice in the case that the map shuffle locations is a `DefaultMapShuffleLocations`.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Details: rootDirs in DiskBlockManagerSuite doesn't get full path from rootDir0, rootDir1