[
https://issues.apache.org/jira/browse/FALCON-1343?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14769006#comment-14769006
]
Balu Vellanki commented on FALCON-1343:
---------------------------------------
[~pallavi.rao] and [~sowmyaramesh] : The validateLocations method throws an
error as follows
{code}
throw new ValidationException(
"Unable to validate the location with path: " + location +
" for cluster:" + clusterName
+ " due to transient failures ", e);
{code}
This error caused confusion to users in the past, especially when submitting
cluster through a UI. They do not have access to full stack trace in the
falcon.application.log. By doing a very simple read/write filesystem
validation, we can solve this problem and show the correct error. This also
catches the error at the right place.
> validation of read/write endpoints is not reliable
> --------------------------------------------------
>
> Key: FALCON-1343
> URL: https://issues.apache.org/jira/browse/FALCON-1343
> Project: Falcon
> Issue Type: Sub-task
> Components: general
> Reporter: Balu Vellanki
> Assignee: Balu Vellanki
> Attachments: FALCON-1343.patch
>
>
> A read/write endpoint is currently validated by creating a filesystem with
> the endpoint url.
> {code}
> HadoopClientFactory.get().createProxiedFileSystem(conf);
> {code}
> I confirmed with a HDFS team member that it is not sufficient validation.
> Ideally check if the end user has access by doing atleast a list /tmp call
> after creating proxiedFileSystem.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
