[
https://issues.apache.org/jira/browse/SPARK-14687?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Liwei Lin updated SPARK-14687:
------------------------------
Description:
Generally we should call path.getFileSystem(conf) instead of call
FileSystem.get(conf), because the latter is actually called on the DEFAULT_URI
(fs.defaultFS), leading to problems under certain situations:
- if {{fs.defaultFS}} is not specified, the schema will default to {{file:///}}
- if {{fs.defaultFS}} is not {{hdfs://...}}, for example {{viewfs://}}(which is
used for federated HDFS)
- if {{fs.defaultFS}} is {{hdfs://A/...}}, but path is {{hdfs://B/...}}
was:
Generally we should call path.getFileSystem(conf) instead of call
FileSystem.get(conf), because the latter is actually called on the DEFAULT_URI
(fs.defaultFS), leading to problems under certain situations:
- if {{fs.defaultFS}} is not specified, the schema will default to {{file:///}}
- if {{fs.defaultFS}} is specified to something other than {{hdfs://}} --
{{viewfs://}}, used for federated HDFS ,as an example
> Call path.getFileSystem(conf) instead of call FileSystem.get(conf)
> ------------------------------------------------------------------
>
> Key: SPARK-14687
> URL: https://issues.apache.org/jira/browse/SPARK-14687
> Project: Spark
> Issue Type: Improvement
> Components: MLlib, Spark Core, SQL
> Affects Versions: 2.0.0
> Reporter: Liwei Lin
> Priority: Minor
>
> Generally we should call path.getFileSystem(conf) instead of call
> FileSystem.get(conf), because the latter is actually called on the
> DEFAULT_URI (fs.defaultFS), leading to problems under certain situations:
> - if {{fs.defaultFS}} is not specified, the schema will default to
> {{file:///}}
> - if {{fs.defaultFS}} is not {{hdfs://...}}, for example {{viewfs://}}(which
> is used for federated HDFS)
> - if {{fs.defaultFS}} is {{hdfs://A/...}}, but path is {{hdfs://B/...}}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]