Todd, can you please hold off for the merge till Tuesday, until some of the other folks working on HA could catch up with some of the recent changes.
On Fri, Feb 3, 2012 at 2:10 PM, Todd Lipcon <t...@cloudera.com> wrote: > I've got a merge pending of trunk into HDFS-1623 -- it was a bit > complicated so wanted to ask for another set of eyes: > https://github.com/toddlipcon/hadoop-common/tree/ha-merge-20120203 > (using github since it's hard to review a merge patch via JIRA) > > The interesting bit of the merge was to deal with conflicts with > HDFS-2718. To summarize the changes I had to make: > - in the HDFS-1623 branch, we don't deal with the case where OP_ADD > contains blocks on a new file -- this is a case that doesn't happen on > real clusters, but currently happens with synthetic logs generated > from the CreateEditLogs tool. I added a TODO to add a sanity check > here and will address as a follow-up. Given the difference between > trunk and branch, there were a couple of small changes that propagated > into unprotectedAddFile > - In the HDFS-1623 branch we had already implemented the > "updateBlocks" call inside FSEditLogLoader. I used that existing > implementation rather than adding the new one in FSDirectory, since > this function had some other changes related to HA in the branch > version. > > I'll wait for a +1 before committing. I ran all of the unit tests and > they passed. > > -Todd > -- > Todd Lipcon > Software Engineer, Cloudera >
