SPARK-2881: Avoid collisisions in Snappy staging directory. #1994
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
By default Snappy uses java.io.tempdir for copying the Snappy native library.
If two users run Spark jobs on the same machine it can cause an exception when
the second user tries to access or overwrite the snappy file created by the
first user. This will fail Spark jobs out-of-the-box if they are run on a
machine shared by different users.
Snappy does expose a mechanism to customize the temp directory via a system
property. This system property is read in a static block inside of Snappy code.
I've added a "best effort" fix for this where we try to set the system
property in a static block before Snappy reads it. I've tested it and it does
work, but it relies on static initialization order which is brittle. I.e.
if user code accesses Snappy libraries first this could not-work.
An alternative work-around for users is to explicitly set
org.xerial.snappy.tempdir
themselves through Spark's java options. I also filed a bug upstream with
Snappy-java to ask them for better behavior here:
xerial/snappy-java#84
I think this is worth merging because in many cases it will fix the issue and
at worst it's a no-op.
pwendell
added a commit
to pwendell/spark
that referenced
this pull request
Aug 17, 2014
This upgrades snappy-java which fixes the issue reported in SPARK-2881. This is the master branch equivalemnt to apache#1994 which provides a different work-around for the 1.1 branch.
This was referenced Aug 17, 2014
Contributor
Author
|
I'm closing this in favor of #1999 |
asfgit
pushed a commit
that referenced
this pull request
Aug 18, 2014
This upgrades snappy-java which fixes the issue reported in SPARK-2881. This is the master branch equivalent to #1994 which provides a different work-around for the 1.1 branch. Author: Patrick Wendell <[email protected]> Closes #1995 from pwendell/snappy-1.1 and squashes the following commits: 0c7c4c2 [Patrick Wendell] SPARK-2881. Upgrade snappy-java to 1.1.1.3.
xiliu82
pushed a commit
to xiliu82/spark
that referenced
this pull request
Sep 4, 2014
This upgrades snappy-java which fixes the issue reported in SPARK-2881. This is the master branch equivalent to apache#1994 which provides a different work-around for the 1.1 branch. Author: Patrick Wendell <[email protected]> Closes apache#1995 from pwendell/snappy-1.1 and squashes the following commits: 0c7c4c2 [Patrick Wendell] SPARK-2881. Upgrade snappy-java to 1.1.1.3.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
By default Snappy uses java.io.tempdir for copying the Snappy native library.
If two users run Spark jobs on the same machine it can cause an exception when
the second user tries to access or overwrite the snappy file created by the
first user. This will fail Spark jobs out-of-the-box if they are run on a
machine shared by different users.
Snappy does expose a mechanism to customize the temp directory via a system
property. This system property is read in a static block inside of Snappy code.
Snappy-java fixes this in newer versions. We can upgrade snappy-java in master
but this proposes a "best effort" fix for 1.1 where we try to set the system
property in a static block before Snappy reads it. I've tested it and it does
work, but it relies on static initialization order which is brittle. I.e.
if user code accesses Snappy libraries first this could not-work.
An alternative work-around for users is to explicitly set
themselves through Spark's java options. I also filed a bug upstream with
Snappy-java to ask them for better behavior here:
xerial/snappy-java#84
I think this is worth merging because in many cases it will fix the issue and
at worst it's a no-op.