[
https://issues.apache.org/jira/browse/HADOOP-14918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16490288#comment-16490288
]
Aaron Fabbri commented on HADOOP-14918:
---------------------------------------
Thanks for making progress on this, [~gabor.bota], it will be great to have
this fixed.
{quote}
I think we should have the caller nominate a specific table to create and
delete, skip the test if unset.
{quote}
+1
Was thinking something similar to [~ste...@apache.org] here. I frequently run
into problems with both {{AbstractS3GuardToolTestBase#testDestroyNoBucket}} and
{{ITestS3GuardToolDynamoDB#testDynamoDBInitDestroyCycle}}. Not only can they
nuke the table out from under other tests, but AWS's handling of table names is
very *eventually* consistent and flaky, meaning a recently re-created table can
occasionally show up as still being deleted.
Should the "destructive s3guard integration test dynamodb table" name go in the
contract test options?
If set, run in serial test phase (see hadoop-aws/pom.xml). If unset, skip
those tests.
> remove the Local Dynamo DB test option
> --------------------------------------
>
> Key: HADOOP-14918
> URL: https://issues.apache.org/jira/browse/HADOOP-14918
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 2.9.0, 3.0.0
> Reporter: Steve Loughran
> Assignee: Gabor Bota
> Priority: Major
> Attachments: HADOOP-14918-001.patch, HADOOP-14918-002.patch,
> HADOOP-14918-003.patch, HADOOP-14918-004.patch
>
>
> I'm going to propose cutting out the localdynamo test option for s3guard
> * the local DDB JAR is unmaintained/lags the SDK We work with...eventually
> there'll be differences in API.
> * as the local dynamo DB is unshaded. it complicates classpath setup for the
> build. Remove it and there's no need to worry about versions of anything
> other than the shaded AWS
> * it complicates test runs. Now we need to test for both localdynamo *and*
> real dynamo
> * but we can't ignore real dynamo, because that's the one which matters
> While the local option promises to reduce test costs, really, it's just
> adding complexity. If you are testing with s3guard, you need to have a real
> table to test against., And with the exception of those people testing s3a
> against non-AWS, consistent endpoints, everyone should be testing with
> S3Guard.
> -Straightforward to remove.-
--
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
