[jira] [Comment Edited] (HADOOP-15420) s3guard ITestS3GuardToolLocal failures in diff tests
[
https://issues.apache.org/jira/browse/HADOOP-15420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16467244#comment-16467244
]
Gabor Bota edited comment on HADOOP-15420 at 5/8/18 11:20 AM:
--
*using {{standardize(Path)}}*
Do you mean using it like:
{code:java}
private boolean expired(FileStatus status, long expiry, String keyPrefix) {
Path path = standardize(status.getPath());
String bucket = path.toUri().getHost();
String translatedPath = "";
if(bucket != null && !bucket.isEmpty()){
translatedPath =
PathMetadataDynamoDBTranslation.pathToParentKey(path);
} else {
translatedPath = path.toString();
}
return status.getModificationTime() < expiry && !status.isDirectory()
&& translatedPath.startsWith(keyPrefix);
}
{code}
I need to check for the bucket. Removing the check for the {{getHost}}
(existing bucket) the following tests will fail:
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneUnsetsAuthoritative
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneFiles
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneDirs
in these cases the {{status.getPath()}}=file:/unpruned-root-dir; and it can not
be digested by {{PathMetadataDynamoDBTranslation.pathToParentKey()}}, the test
will fail with the following:
{noformat}
java.lang.IllegalArgumentException: Path missing bucket
at
com.google.common.base.Preconditions.checkArgument(Preconditions.java:88)
at
org.apache.hadoop.fs.s3a.s3guard.PathMetadataDynamoDBTranslation.pathToParentKey(PathMetadataDynamoDBTranslation.java:255)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.expired(LocalMetadataStore.java:358)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:318)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:308)
at
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase.testPruneUnsetsAuthoritative(MetadataStoreTestBase.java:730)
(...)
{noformat}
*Testing with local dynamo*
Here is my output for {{mvn clean test -Ds3guard -Ddynamo}}:
[https://pastebin.com/raw/qGuwuS4F]
Short summary: Tests run: 398, Failures: 0, Errors: 0, Skipped: 2
was (Author: gabor.bota):
*using {{standardize(Path)}}*
Do you mean using it like:
{code:java}
private boolean expired(FileStatus status, long expiry, String keyPrefix) {
Path path = standardize(status.getPath());
String bucket = path.toUri().getHost();
String translatedPath = "";
if(bucket != null && !bucket.isEmpty()){
translatedPath =
PathMetadataDynamoDBTranslation.pathToParentKey(path);
} else {
translatedPath = path.toString();
}
return status.getModificationTime() < expiry && !status.isDirectory()
&& translatedPath.startsWith(keyPrefix);
}
{code}
I need to check for the bucket. Removing the check for the {{getHost}}
(existing bucket) the following tests will fail:
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneUnsetsAuthoritative
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneFiles
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneDirs
in these cases the {{status.getPath()}}=file:/unpruned-root-dir; and it can not
be digested by {{PathMetadataDynamoDBTranslation.pathToParentKey()}}, the test
will fail with the following:
{noformat}
java.lang.IllegalArgumentException: Path missing bucket
at
com.google.common.base.Preconditions.checkArgument(Preconditions.java:88)
at
org.apache.hadoop.fs.s3a.s3guard.PathMetadataDynamoDBTranslation.pathToParentKey(PathMetadataDynamoDBTranslation.java:255)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.expired(LocalMetadataStore.java:358)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:318)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:308)
at
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase.testPruneUnsetsAuthoritative(MetadataStoreTestBase.java:730)
(...)
{noformat}
*Testing with local dynamo*
Here is my output for {{mvn clean test -Ds3guard -Ddynamo}}:
[https://pastebin.com/qGuwuS4F]
Short summary: Tests run: 398, Failures: 0, Errors: 0, Skipped: 2
> s3guard ITestS3GuardToolLocal failures in diff tests
>
>
> Key: HADOOP-15420
> URL: https://issues.apache.org/jira/browse/HADOOP-15420
> Project: Hadoop Common
> Issue Type: Sub-task
>Reporter: Aaron Fabbri
>Assignee: Gabor Bota
>Priority: Minor
> Attachments: HADOOP-15420.001.patch, HADOOP-15420.002.patch
>
>
> Noticed this when testing the patch for HADOOP-13756.
>
> {code:java}
> [ERROR]
[jira] [Comment Edited] (HADOOP-15420) s3guard ITestS3GuardToolLocal failures in diff tests
[
https://issues.apache.org/jira/browse/HADOOP-15420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16467244#comment-16467244
]
Gabor Bota edited comment on HADOOP-15420 at 5/8/18 11:16 AM:
--
*using {{standardize(Path)}}*
Do you mean using it like:
{code:java}
private boolean expired(FileStatus status, long expiry, String keyPrefix) {
Path path = standardize(status.getPath());
String bucket = path.toUri().getHost();
String translatedPath = "";
if(bucket != null && !bucket.isEmpty()){
translatedPath =
PathMetadataDynamoDBTranslation.pathToParentKey(path);
} else {
translatedPath = path.toString();
}
return status.getModificationTime() < expiry && !status.isDirectory()
&& translatedPath.startsWith(keyPrefix);
}
{code}
I need to check for the bucket. Removing the check for the {{getHost}}
(existing bucket) the following tests will fail:
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneUnsetsAuthoritative
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneFiles
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneDirs
in these cases the {{status.getPath()}}=file:/unpruned-root-dir; and it can not
be digested by {{PathMetadataDynamoDBTranslation.pathToParentKey()}}, the test
will fail with the following:
{noformat}
java.lang.IllegalArgumentException: Path missing bucket
at
com.google.common.base.Preconditions.checkArgument(Preconditions.java:88)
at
org.apache.hadoop.fs.s3a.s3guard.PathMetadataDynamoDBTranslation.pathToParentKey(PathMetadataDynamoDBTranslation.java:255)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.expired(LocalMetadataStore.java:358)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:318)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:308)
at
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase.testPruneUnsetsAuthoritative(MetadataStoreTestBase.java:730)
(...)
{noformat}
*Testing with local dynamo*
Here is my output for {{mvn clean test -Ds3guard -Ddynamo}}:
[https://pastebin.com/qGuwuS4F]
Short summary: Tests run: 398, Failures: 0, Errors: 0, Skipped: 2
was (Author: gabor.bota):
*using {{standardize(Path)}}
Do you mean using it like:
{code:java}
private boolean expired(FileStatus status, long expiry, String keyPrefix) {
Path path = standardize(status.getPath());
String bucket = path.toUri().getHost();
String translatedPath = "";
if(bucket != null && !bucket.isEmpty()){
translatedPath =
PathMetadataDynamoDBTranslation.pathToParentKey(path);
} else {
translatedPath = path.toString();
}
return status.getModificationTime() < expiry && !status.isDirectory()
&& translatedPath.startsWith(keyPrefix);
}
{code}
I need to check for the bucket. Removing the check for the {{getHost}}
(existing bucket) the following tests will fail:
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneUnsetsAuthoritative
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneFiles
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase#testPruneDirs
in these cases the {{status.getPath()}}=file:/unpruned-root-dir; and it can not
be digested by {{PathMetadataDynamoDBTranslation.pathToParentKey()}}, the test
will fail with the following:
{noformat}
java.lang.IllegalArgumentException: Path missing bucket
at
com.google.common.base.Preconditions.checkArgument(Preconditions.java:88)
at
org.apache.hadoop.fs.s3a.s3guard.PathMetadataDynamoDBTranslation.pathToParentKey(PathMetadataDynamoDBTranslation.java:255)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.expired(LocalMetadataStore.java:358)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:318)
at
org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore.prune(LocalMetadataStore.java:308)
at
org.apache.hadoop.fs.s3a.s3guard.MetadataStoreTestBase.testPruneUnsetsAuthoritative(MetadataStoreTestBase.java:730)
(...)
{noformat}
*Testing with local dynamo*
Here is my output for {{mvn clean test -Ds3guard -Ddynamo}}:
[https://pastebin.com/qGuwuS4F]
Short summary: Tests run: 398, Failures: 0, Errors: 0, Skipped: 2
> s3guard ITestS3GuardToolLocal failures in diff tests
>
>
> Key: HADOOP-15420
> URL: https://issues.apache.org/jira/browse/HADOOP-15420
> Project: Hadoop Common
> Issue Type: Sub-task
>Reporter: Aaron Fabbri
>Assignee: Gabor Bota
>Priority: Minor
> Attachments: HADOOP-15420.001.patch, HADOOP-15420.002.patch
>
>
> Noticed this when testing the patch for HADOOP-13756.
>
> {code:java}
> [ERROR]
[jira] [Comment Edited] (HADOOP-15420) s3guard ITestS3GuardToolLocal failures in diff tests
[
https://issues.apache.org/jira/browse/HADOOP-15420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16466764#comment-16466764
]
Aaron Fabbri edited comment on HADOOP-15420 at 5/8/18 2:44 AM:
---
Thank you for working on this issue [~gabor.bota]. Good work identifying the
bug. Couple comments on v1 patch:
{noformat}
private boolean expired(FileStatus status, long expiry, String keyPrefix) {
+// remove the protocol from path string to be able to compare
+String bucket = status.getPath().toUri().getHost();
+ statusTranslatedPath = status.getPath().toUri().getPath();
+}
+
{noformat}
Can you use helper func {{standardize(Path)}} here instead?
Thanks for moving {{ testDiffCommand() }} to the base class. Did you test this
with Dynamo? (`mvn clean test -Ds3guard -Ddynamo`) Unfortunately dynamodb
metadatastore test the Local (in-memory) Test Dynamo thing still (until we
finish HADOOP-14918).
was (Author: fabbri):
Thank you for working on this issue [~gabor.bota]. Good work identifying the
bug. Couple comments on v1 patch:
{noformat}
private boolean expired(FileStatus status, long expiry, String keyPrefix) {
+// remove the protocol from path string to be able to compare
+String bucket = status.getPath().toUri().getHost();
+ statusTranslatedPath = status.getPath().toUri().getPath();
+}
+
{noformat}
Can you use helper func {{standardize(Path)}} here instead?
Thanks for moving {{ testDiffCommand() }} to the base class. Did you test this
with Dynamo? (`mvn clean test -Ds3guard -Ddynamo`) Unfortunately dynamodb
metadatastore test the Local (in-memory) Test Dynamo thing still (until we
finish HADOOP-14918).
Also a reminder please declare which AWS region you ran integration tests in.
> s3guard ITestS3GuardToolLocal failures in diff tests
>
>
> Key: HADOOP-15420
> URL: https://issues.apache.org/jira/browse/HADOOP-15420
> Project: Hadoop Common
> Issue Type: Sub-task
>Reporter: Aaron Fabbri
>Assignee: Gabor Bota
>Priority: Minor
> Attachments: HADOOP-15420.001.patch, HADOOP-15420.002.patch
>
>
> Noticed this when testing the patch for HADOOP-13756.
>
> {code:java}
> [ERROR] Failures:
> [ERROR]
> ITestS3GuardToolLocal>AbstractS3GuardToolTestBase.testPruneCommandCLI:221->AbstractS3GuardToolTestBase.testPruneCommand:201->AbstractS3GuardToolTestBase.assertMetastoreListingCount:214->Assert.assertEquals:555->Assert.assertEquals:118->Assert.failNotEquals:743->Assert.fail:88
> Pruned children count
> [PathMetadata{fileStatus=S3AFileStatus{path=s3a://bucket-new/test/testPruneCommandCLI/stale;
> isDirectory=false; length=100; replication=1; blocksize=512;
> modification_time=1524798258286; access_time=0; owner=hdfs; group=hdfs;
> permission=rw-rw-rw-; isSymlink=false; hasAcl=false; isEncrypted=false;
> isErasureCoded=false} isEmptyDirectory=FALSE; isEmptyDirectory=UNKNOWN;
> isDeleted=false},
> PathMetadata{fileStatus=S3AFileStatus{path=s3a://bucket-new/test/testPruneCommandCLI/fresh;
> isDirectory=false; length=100; replication=1; blocksize=512;
> modification_time=1524798262583; access_time=0; owner=hdfs; group=hdfs;
> permission=rw-rw-rw-; isSymlink=false; hasAcl=false; isEncrypted=false;
> isErasureCoded=false} isEmptyDirectory=FALSE; isEmptyDirectory=UNKNOWN;
> isDeleted=false}] expected:<1> but was:<2>{code}
>
> Looking through the code, I'm noticing a couple of issues.
>
> 1. {{testDiffCommand()}} is in {{ITestS3GuardToolLocal}}, but it should
> really be running for all MetadataStore implementations. Seems like it
> should live in {{AbstractS3GuardToolTestBase}}.
> 2. {{AbstractS3GuardToolTestBase#createFile()}} seems wrong. When
> {{onMetadataStore}} is false, it does a {{ContractTestUtils.touch(file)}},
> but the fs is initialized with a MetadataStore present, so seem like the fs
> will still put the file in the MetadataStore?
> There are other tests which explicitly go around the MetadataStore by using
> {{fs.setMetadataStore(nullMS)}}, e.g. ITestS3AInconsistency. We should do
> something similar in {{AbstractS3GuardToolTestBase#createFile()}}, minding
> any issues with parallel test runs.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
