Hi all,
I've created a branch in SVN called HDFS-3042. I plan on working on
the ZK-based automatic failover solution (HDFS-3042, design doc on
HDFS-2185) in this branch. We'll continue to use the normal JIRA
process, with review-then-commit policies in effect.
The reason for the branch is that we'd like to move quickly to a
working implementation, but there are still some open questions on the
finer points of the design. Some folks have expressed concerns about
moving too quickly on this project on trunk, whereas working on a
feature branch will let us take a bit more iterative of an approach. I
think this will enable us to converge more quickly to a stable
implementation of automatic failover in the coming weeks.
Given that I already have automatic failover working in my test
environment, I don't imagine this branch will have to live very long
-- if all goes well, on the order of a week or two.
As with all such branches, we'll get a green test build and take a
vote before merging back into trunk -- a merge of a feature branch
requires 3 committer +1s.
Of course, the branch is open to all to contribute! I've also created
a fix version ("Auto failover (HDFS-3042)") in HDFS and Common JIRAs,
and an "auto-failover" component. I have not yet assigned these tags
to the open JIRAs but will do so tomorrow.
Thanks
-Todd
--
Todd Lipcon
Software Engineer, Cloudera
