[
https://issues.apache.org/jira/browse/HADOOP-13972?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15818304#comment-15818304
]
Steve Loughran commented on HADOOP-13972:
-----------------------------------------
I propose you follow the pattern of HADOOP-13336, where every property
{{fs.s3a.bucket.$BUCKET.field1.field2=value}} is patched in to
{{fs.s3a.field1.field2=}} for the FS named $BUCKET. This allows for every FS
property to be overridden, rather than per-store keys.
# This structure follows that of the hdfs HA nn property settings.
# It's easy to add; you don't need to implement the complex "read one then
fallback" strategy, just do a patch in FileSystem.init().
I'm about the put the patch into branch-2.8+ with the core code in S3AUtils.
Once it's in, you could pull it up into hadoop common and make more generic,
that way we'd have a consistent implementation across filesystems.
> Access multiple Azure Data Lake stores with different SPIs
> ----------------------------------------------------------
>
> Key: HADOOP-13972
> URL: https://issues.apache.org/jira/browse/HADOOP-13972
> Project: Hadoop Common
> Issue Type: Improvement
> Components: fs/adl
> Affects Versions: 3.0.0-alpha2
> Reporter: John Zhuge
>
> Useful when distcp needs to access 2 Data Lake stores with different SPIs.
> Of course, a workaround is to grant the same SPI access permission to both
> stores, but sometimes it might not be feasible.
> One idea is to embed the store name in the configuration property names,
> e.g., {{dfs.adls.oauth2.<store>.client.id}}. Per-store keys will be consulted
> first, then fall back to the global keys.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
