S3 file systems should not create bucket

• Key: HADOOP-4422
• Type: Bug
• Status: Closed
• Resolution: Fixed
• Priority: Major
• Assignee: David Phillips
• Reporter: David Phillips
• Component/s: fs/s3
• Affects Version/s: 0.18.1
• Fix Version/s: 0.20.0

Description

Both S3 file systems (s3 and s3n) try to create the bucket at every initialization. This is bad because

• Every S3 operation costs money. These unnecessary calls are an unnecessary expense.
• These calls can fail when called concurrently. This makes the file system unusable in large jobs.
• Any operation, such as a "fs -ls", creates a bucket. This is counter-intuitive and undesirable.

The initialization code should assume the bucket exists:

• Creating a bucket is a very rare operation. Accounts are limited to 100 buckets.
• Any check at initialization for bucket existence is a waste of money.

Per Amazon: "Because bucket operations work against a centralized, global resource space, it is not appropriate to make bucket create or delete calls on the high availability code path of your application. It is better to create or delete buckets in a separate initialization or setup routine that you run less often."

Comments

[David Phillips]

Simple patch that removes bucket creation.

[Hadoop QA]

-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12392205/hadoop-s3n-nocreate.patch
against trunk revision 705073.

+1 @author. The patch does not contain any @author tags.

-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no tests are needed for this patch.

-1 patch. The patch command could not apply the patch.

Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3469/console

This message is automatically generated.

[Doug Cutting]

This patch needs to be re-generated without the a/ b/ stuff for Hudson to be able to apply it. It must apply with 'patch -p 0 < foo.patch' when connected to trunk.

[David Phillips]

Patch applies with -p0 now (used git diff --no-prefix).

[Tom White]

Marked as an incompatible change, since existing code that relies on bucket creation will need to be changed.

[Hadoop QA]

-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12392277/hadoop-s3n-nocreate.patch
against trunk revision 705430.

+1 @author. The patch does not contain any @author tags.

-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no tests are needed for this patch.

+1 javadoc. The javadoc tool did not generate any warning messages.

+1 javac. The applied patch does not increase the total number of javac compiler warnings.

+1 findbugs. The patch does not introduce any new Findbugs warnings.

+1 Eclipse classpath. The patch retains Eclipse classpath integrity.

+1 core tests. The patch passed core unit tests.

+1 contrib tests. The patch passed contrib unit tests.

Test results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3482/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3482/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3482/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3482/console

This message is automatically generated.

[Tom White]

For consistency, we should make the same change to Jets3tFileSystemStore.

Also, regarding the tests, there are two unit tests: Jets3tS3FileSystemContractTest and Jets3tNativeS3FileSystemContractTest which can be run manually to test the S3 integration. The only difference with this patch is that the buckets they run against must already exist - so I don't think any change to the tests are needed.

[Tom White]

Cancelling patch pending change to Jets3tFileSystemStore.

[David Phillips]

Bucket creation also removed from Jets3tFileSystemStore.

[David Phillips]

I ran the tests as follows after setting the correct test buckets and keys in src/test/hadoop-site.xml:

ant -Dtestcase=Jets3tS3FileSystemContractTest test
ant -Dtestcase=Jets3tNativeS3FileSystemContractTest test

They seem to pass:

Testsuite: org.apache.hadoop.fs.s3.Jets3tS3FileSystemContractTest
Tests run: 25, Failures: 0, Errors: 0, Time elapsed: 131.575 sec
Testsuite: org.apache.hadoop.fs.s3native.Jets3tNativeS3FileSystemContractTest
Tests run: 26, Failures: 0, Errors: 0, Time elapsed: 52.694 sec

However, they both produce hundreds of warnings:

(s3) 2008-11-21 15:51:28,800 WARN httpclient.RestS3Service (RestS3Service.java:performRequest(317)) - Response '/%2Ftest' - Unexpected response code 404, expected 200
(s3n) 2008-11-21 15:34:55,646 WARN httpclient.RestS3Service (RestS3Service.java:performRequest(317)) - Response '/test' - Unexpected response code 404, expected 200

Any ideas?

[David Phillips]

Never mind about the warnings from Jets3t during testing. They are expected.

[Hadoop QA]

-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12394458/hadoop-s3n-nocreate.patch
against trunk revision 719787.

+1 @author. The patch does not contain any @author tags.

-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no tests are needed for this patch.

+1 javadoc. The javadoc tool did not generate any warning messages.

+1 javac. The applied patch does not increase the total number of javac compiler warnings.

+1 findbugs. The patch does not introduce any new Findbugs warnings.

+1 Eclipse classpath. The patch retains Eclipse classpath integrity.

+1 core tests. The patch passed core unit tests.

+1 contrib tests. The patch passed contrib unit tests.

Test results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3638/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3638/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3638/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3638/console

This message is automatically generated.

[Tom White]

I've just committed this. Thanks David!

[Hudson]

Integrated in Hadoop-trunk #670 (See http://hudson.zones.apache.org/hudson/job/Hadoop-trunk/670/)
. S3 file systems should not create bucket. Contributed by David Phillips.

[Robert Chansler]

Edit release note for publication.