[
https://issues.apache.org/jira/browse/MAPREDUCE-5809?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13947252#comment-13947252
]
Chris Nauroth commented on MAPREDUCE-5809:
------------------------------------------
[~sureshms], yes, this is a very important consideration. Here is what I had
in mind for the logic:
# distcp CLI accepts a new optional flag: -pa for "preserve ACLs". The
presence of -pa also implies the existing -pp flag, because ACLs are a
super-set of permissions.
# If preserving ACLs, then before submitting the job, distcp sends a canary
{{getAclStatus}} request for / on the source and target file systems. This
will detect ACL compatibility/support problems and fail fast before even
submitting the job. There are three specific sub-cases that this check catches:
## File system is HDFS < 2.4, so the getAclStatus RPC endpoint doesn't exist.
## File system is HDFS >= 2.4, but ACLs are not enabled.
## File system is a {{FileSystem}} subclass that doesn't override the ACL APIs.
({{UnsupportedOperationException}})
# Then, distcp map tasks call {{getAclStatus}} and {{setAcl}} instead of
{{setPermission}}.
If the operator wishes to copy files with ACLs to a target cluster and just
drop the ACLs, then they'd run without the -pa option, and distcp would
continue to work as it does today.
One additional thing I just learned while reading the code is that we support
passing the -p flag with no additional arguments, and this is assumed to
preserve replication, block size, user, group, permission and checksum type.
I'm planning on preserving this behavior. I don't think we can include -pa as
part of the defaults, because that could break existing deployments that are
running distcp -p if they start using ACLs on one cluster but not the other.
> Enhance distcp to support preserving HDFS ACLs.
> -----------------------------------------------
>
> Key: MAPREDUCE-5809
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5809
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: distcp
> Affects Versions: 2.4.0
> Reporter: Chris Nauroth
> Assignee: Chris Nauroth
>
> This issue tracks enhancing distcp to add a new command-line argument for
> preserving HDFS ACLs from the source at the copy destination.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
