I'm generally supportive of a merge - unless anyone brings up new concerns, I've been happy with the design, implementation and discussion and it seems to be at a good checkpoint.
When you get to writing the site documentation - one thing in particular I think should be called out early in the description of the feature is the need to create a read-only namespace specific to the remote storage. I can see this getting merged and more than a few people wanting to know how they can dynamically mount their remote storage on their existing HDFS cluster today :) On Sun, Dec 3, 2017 at 3:43 PM, Chris Douglas <cdoug...@apache.org> wrote: > Hey hdfs-dev@- > > The HDFS-9806 dev branch is getting ready to merge to trunk. This branch > adds a new storage type (PROVIDED) to support reading remote data as HDFS > blocks. Design documentation and discussion are available on JIRA [1]. > > Over the next week or so, we'll work through the remaining issues, write > the site documentation, and generally clean up the branch. The current code > has been stressed in experimental settings, including a large production > benchmark. When this feature is not enabled and active, it should have no > effect on HDFS. The merge will checkpoint progress on reading remote data > through HDFS, setting up work on the write path [2] and improving > usability. > > Please look over the branch and raise issues that should be addressed > before the merge. -C > > [1]: https://issues.apache.org/jira/browse/HDFS-9806 > [2]: https://issues.apache.org/jira/browse/HDFS-12090 >
