[
https://issues.apache.org/jira/browse/HADOOP-13452?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15415573#comment-15415573
]
Aaron Fabbri commented on HADOOP-13452:
---------------------------------------
Hi [~cnauroth]. I'm happy to provide code reviews for HADOOP-13447.
I was just thinking that HADOOP-13448 is also a big prerequisite for a lot of
this work. That interface definition will be a good boundary for splitting up
work.
If we define that interface such that each side (MetadataStore impl., and
clients that use it) can decide on policy, then your suggestion makes sense for
the S3A client side behavior. I.e., if the access policy in S3A is set to
allow the MetadataStore as source of truth, it will push full directory
contents with "fully cached" flag into the MetadataStore (MS). It would also
honor the "fully cached" semantics on MS lookup. Those details would change
depending on the MS design we settle on, but the example should make sense.
> S3Guard: Implement access policy for intra-client consistency with in-memory
> metadata store.
> --------------------------------------------------------------------------------------------
>
> Key: HADOOP-13452
> URL: https://issues.apache.org/jira/browse/HADOOP-13452
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Reporter: Chris Nauroth
> Assignee: Aaron Fabbri
>
> Implement an S3A access policy based on an in-memory metadata store. This
> can provide consistency within the same client without needing to integrate
> with an external system.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
