[
https://issues.apache.org/jira/browse/HDDS-4097?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17186412#comment-17186412
]
Marton Elek commented on HDDS-4097:
-----------------------------------
Well, that's a valid question, but I think it might be possible to support it
on some level.
Let's say we create implicit dir entries with a flag (impicit=?)
Create Key /a/b/c/file1
Created real keys:
/a (implicit=true)
/a/b (implicit=true)
/a/b/c (implicit=true)
/a/b/c/file1 (implicit=false)
In this case you can create dir (!) entry a/b/c, but we need to make /a/b/c
(implicit=false). We can also hide all the implicit entries by default when
keys are listed from s3 endpoint.
Storage can be the same but o3fs/ofs and s3g can display different
representation.
The harder question is when /a/b/c/file1 and /a/b/c files are created, both
with content. As far as I see there are multiple options:
1. Reject it from s3 interface (AWS incompatible)
2. Persist it to the keys but show only one from hcfs interface (eg. show only
the file without subkeys / dirs, or show only the dirs)
3. We can also show a technical file on hcfs interface (like .CONFLICT) in
additional to the dir which can solve this problem
And we have a similar question about the normalization
I am not sure what is the best decision here (1, 2, 3 or other). But I have
some concerns that we're giving up from the compatibility without collection
all the incompatibilities which will be introduced, or without checking what
are the minimum level of the required compatibility support.
They might be collected and considered but I miss it.
In general, I think trying to be S3 compatible is a very important goal (as we
are an object store), but also understand that providing file system semantics
is also important. What I am interested is the support of both, with just the
minimal level of differences with S3.
> S3/Ozone Filesystem inter-op
> ----------------------------
>
> Key: HDDS-4097
> URL: https://issues.apache.org/jira/browse/HDDS-4097
> Project: Hadoop Distributed Data Store
> Issue Type: New Feature
> Reporter: Bharat Viswanadham
> Assignee: Bharat Viswanadham
> Priority: Major
> Attachments: Ozone FileSystem Paths Enabled.docx, Ozone filesystem
> path enabled.xlsx
>
>
> This Jira is to implement changes required to use Ozone buckets when data is
> ingested via S3 and use the bucket/volume via OzoneFileSystem. Initial
> implementation for this is done as part of HDDS-3955. There are few API's
> which have missed the changes during the implementation of HDDS-3955.
> Attached design document which discusses each API, and what changes are
> required.
> Excel sheet has information about each API, from what all interfaces the OM
> API is used, and what changes are required for the API to support
> inter-operability.
> Note: The proposal for delete/rename is still under discussion, not yet
> finalized.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: ozone-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: ozone-issues-h...@hadoop.apache.org
