[
https://issues.apache.org/jira/browse/HADOOP-12548?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14989390#comment-14989390
]
Steve Loughran commented on HADOOP-12548:
-----------------------------------------
S3a can also talk to Amazon's EC2 metadata service to fetch them off EC2
infrastructure ... this is probably only useful on EC2 nodes. You may be able
to use other {{AWSCredentialsProvider}} subclasses, such as amazon's own
{{PropertiesFileCredentialsProvider}} and
{{ClasspathPropertiesFileCredentialsProvider}}.
# s3a is the one where this stuff should take place
# & yes, it should be the CredentialsProvider to ask for a password. And an
account
# allowing clients to select specific {{AWSCredentialsProvider}}
implementations by classname could be very extensible.
# if you are going to do this, why not add it to sf/swift too?
One recurrent "how do I" question is "how do I support multiple s3 buckets with
different accounts? The swift code does it by swift endpoint; there's nothing
in the s3 code. Would it be possible to somehow use the bucket name when
looking up a username/password? Maybe do a first attempt to lookup
{{fs.s3a.${BUCKET}.access.key}} & same for password, falling back to the base
one if unset?
> read s3 creds from a file
> -------------------------
>
> Key: HADOOP-12548
> URL: https://issues.apache.org/jira/browse/HADOOP-12548
> Project: Hadoop Common
> Issue Type: New Feature
> Components: fs/s3
> Reporter: Allen Wittenauer
>
> It would be good if we could read s3 creds from a file rather than via a java
> property.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
