[jira] [Updated] (HADOOP-13502) Rename/split fs.contract.is-blobstore flag used by contract tests.

["Chris Nauroth (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HADOOP-13502?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Chris Nauroth updated HADOOP-13502:
-----------------------------------
    Attachment: HADOOP-13502-branch-2.001.patch

I'm attaching patch 001.  Summary:

* Introduced 2 new specific contract options: create-overwrites-directory and create-visibility-delayed.
* Deprecated the is-blobstore flag, but retained it in case file system implementations outside the Hadoop source tree are using it.  (Side note: do we need to add audience and stability annotations to the contract test classes?)
* Cleaned up a few minor JavaDoc omissions in {{ContractOptions}}.
* Updated contract XML configuration files to remove usage of is-blobstore for the actively maintained file systems.  Notice that S3A adds create-visibility-delayed, but does not add create-overwrites-directory.  That's because HADOOP-13188 recently changed the implementation to avoid overwriting directories.

So far, I have tested by running all subclasses of {{AbstractContractCreateTest}}, except {{TestSwiftContractCreate}}.  Maybe this will be the patch where I finally get myself set up to run Swift tests.


> Rename/split fs.contract.is-blobstore flag used by contract tests.
> ------------------------------------------------------------------
>
>                 Key: HADOOP-13502
>                 URL: https://issues.apache.org/jira/browse/HADOOP-13502
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: test
>            Reporter: Chris Nauroth
>            Assignee: Chris Nauroth
>            Priority: Minor
>         Attachments: HADOOP-13502-branch-2.001.patch
>
>
> The {{fs.contract.is-blobstore}} flag guards against execution of several contract tests to account for known limitations with blob stores.  However, the name is not entirely accurate, because it's still possible that a file system implemented against a blob store could pass those tests, depending on whether or not the implementation matches the semantics of HDFS.  This issue proposes to rename the flag or split it into different flags with different definitions for the semantics covered by the current flag.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org