Mingliang Liu commented on HADOOP-14749:

provision burden?
I was thinking that, suppose a user has dozens of buckets in a region, if each 
bucket has a dedicated DDB table, then the user will have to provision dozens 
of tables according to each table's peak/idle load. Instead, if she shares the 
metadata in a single DDB table for all the buckets in that region, she will 
need to only provision one table capacity according to overall usage. This 
amortizes the provision burden.

how to handle s3guard as a read-only client.
I know IAM role can have fine granularity about READ access to S3 bucket (e.g. 
"s3:GetObject") and DDB table (e.g. "dynamodb:Query", "dynamodb:Get" etc). This 
might be considered/operated by the user. But in code, we should not populate 
the metadata from S3 to DDB in the read-only case.

> review s3guard docs & code prior to merge
> -----------------------------------------
>                 Key: HADOOP-14749
>                 URL: https://issues.apache.org/jira/browse/HADOOP-14749
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: documentation, fs/s3
>    Affects Versions: HADOOP-13345
>            Reporter: Steve Loughran
>            Assignee: Steve Loughran
>         Attachments: HADOOP-14749-HADOOP-13345-001.patch, 
> HADOOP-14749-HADOOP-13345-002.patch, HADOOP-14749-HADOOP-13345-003.patch
>   Original Estimate: 24h
>  Remaining Estimate: 24h
> Pre-merge cleanup while it's still easy to do
> * Read through all the docs, tune
> * Diff the trunk/branch files to see if we can reduce the delta (and hence 
> the changes)
> * Review the new tests

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org

Reply via email to