[
https://issues.apache.org/jira/browse/HDFS-12292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16133547#comment-16133547
]
Mikhail Erofeev edited comment on HDFS-12292 at 8/18/17 8:14 PM:
-----------------------------------------------------------------
Hi [~vagarychen], thanks for pointing out to this.
Indeed, my scope was too broad, I moved a call to resolveFsPath() from Command
to DFSAdmin.DFSAdminCommand.
I also added an optimisation check for the same fs in resolveFsPath(), though
for now all paths are normalized from expandAsGlob() and don't benefit from it.
As of you original question viewfs/non-viewfs, I fixed almost as you suggested,
but I checked for DistributedFileSystem instead of just ViewFileSystem, as
other wrapper-fs (ChRootedFileSystem, FilterFileSystem, HarFileSystem) are also
can fail with this exception.
After these fixes TestHDFSCLI no longer needs changes as was in the previous
version!
Thanks again.
was (Author: erofeev):
Hi [~cliang], thanks for pointing out to this.
Indeed, my scope was too broad, I moved a call to resolveFsPath() from Command
to DFSAdmin.DFSAdminCommand.
I also added an optimisation check for the same fs in resolveFsPath(), though
for now all paths are normalized from expandAsGlob() and don't benefit from it.
As of you original question viewfs/non-viewfs, I fixed almost as you suggested,
but I checked for DistributedFileSystem instead of just ViewFileSystem, as
other wrapper-fs (ChRootedFileSystem, FilterFileSystem, HarFileSystem) are also
can fail with this exception.
After these fixes TestHDFSCLI no longer needs changes as was in the previous
version!
Thanks again.
> Federation: Support viewfs:// schema path for DfsAdmin commands
> ---------------------------------------------------------------
>
> Key: HDFS-12292
> URL: https://issues.apache.org/jira/browse/HDFS-12292
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: federation
> Reporter: Mikhail Erofeev
> Assignee: Mikhail Erofeev
> Attachments: HDFS-12292-002.patch, HDFS-12292-003.patch,
> HDFS-12292-004.patch, HDFS-12292.patch
>
>
> Motivation:
> As of now, clients need to specify a nameservice when a cluster is federated,
> otherwise, the exception is fired:
> {code}
> hdfs dfsadmin -setQuota 10 viewfs://vfs-root/user/uname
> setQuota: FileSystem viewfs://vfs-root/ is not an HDFS file system
> # with fs.defaultFS = viewfs://vfs-root/
> hdfs dfsadmin -setQuota 10 vfs-root/user/uname
> setQuota: FileSystem viewfs://vfs-root/ is not an HDFS file system
> # works fine thanks to https://issues.apache.org/jira/browse/HDFS-11432
> hdfs dfsadmin -setQuota 10 hdfs://users-fs/user/uname
> {code}
> This creates inconvenience, inability to rely on fs.defaultFS and forces to
> create client-side mappings for management scripts
> Implementation:
> PathData that is passed to commands should be resolved to its actual
> FileSystem
> Result:
> ViewFS will be resolved to the actual HDFS file system
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
