[ https://issues.apache.org/jira/browse/HADOOP-14971?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16225277#comment-16225277 ]
Steve Loughran commented on HADOOP-14971: ----------------------------------------- Had a conf call w/ Aaron on friday. He will have some fuller comments to make; I've got some (from memory) notes. * Proposed: restore different config options for s3a retry policy and aws SDK policy, so that if s3 wraps AWS it doesn't double the retries. Issue: we aren't 100% sure when AWS sdk retries, though we know that it does for PUT of files, and (in the past) does not for the final POST of an MPU. Action: restore the options. * Some discussion about how to react to NPEs &c in the AWS sdk. For now, do not attempt to retry. * What should we do about delete and idempotency? This is a long standing topic of past HDFS discussion (HDFS-4872). Aaron: what did we conclude here? * Sometimes DDB throws resource not found, esp. on test setup after a new table is created. We way want to treat the DDB exception as retriable, even though S3 404 -> FNFE -> unrecoverable. (steve says: good idea, let's do it after the merge) * Security: we should document the security implications for the committers & have it reviewed. Things to worry about: Denial of service, destroying/tampering with data by altering .pendingset files; aborting pending uploads. The Staging committers use the shared HDFS FS, so need to have security policy there reviewed and documented. I'm reviewing how we use & document this path so that it defaults to being something under the user's home directory. That way: if HDFS is locked down, the path should be too > Merge S3A committers into trunk > ------------------------------- > > Key: HADOOP-14971 > URL: https://issues.apache.org/jira/browse/HADOOP-14971 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 > Affects Versions: 3.0.0 > Reporter: Steve Loughran > Assignee: Steve Loughran > > Merge the HADOOP-13786 committer into trunk. This branch is being set up as a > github PR for review there & to keep it out the mailboxes of the watchers on > the main JIRA -- This message was sent by Atlassian JIRA (v6.4.14#64029) --------------------------------------------------------------------- To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org