Steve Loughran created HADOOP-17559:
---------------------------------------
Summary: S3Guard import can OOM on large imports
Key: HADOOP-17559
URL: https://issues.apache.org/jira/browse/HADOOP-17559
Project: Hadoop Common
Issue Type: Sub-task
Components: fs/s3
Affects Versions: 3.3.1
Reporter: Steve Loughran
Assignee: Steve Loughran
I know I'm closing ~all S3Guard issues as wontfix, but this is pressing so I'm
going to do it anyway
S3guard import of directory tree containing many, many files will OOM. Looking
at the code this is going to be because
* import tool builds a map of all dirs imported, which as the comments note
"superfluous for DDB". - *cut*
* DDB AncestorState tracks files as well as dirs, purely as a safety check to
make sure current op doesn't somehow write a file entry above a dir entry in
the same operation
We've been running S3Guard for a long time, and condition #2 has never arisen.
Propose: don't store filenames there, so memory consumption goes from O(files +
dirs) to O(dirs)
Code straightforward, can't think of any tests
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-dev-h...@hadoop.apache.org
