On Fri, Oct 21, 2011 at 3:40 PM, Matt Foley <[EMAIL PROTECTED]> wrote:
> Hello Roman,
> HDFS-1943 is a resolved bug which was opened only against 0.22 and 0.23. No
> one in the community (including you) saw fit to raise it as a concern for
> 0.20-security, despite multiple public invitations to propose jiras for
> inclusion in 0.20.205.
Well, this is slightly incorrect. It wasn't really fixed for .22.
That's why I opened
https://issues.apache.org/jira/browse/HDFS-2377 as part of Bigtop work on .22.
I hoped that RMs would actively monitor HDFS, MAPRED and COMMON
jira traffic, since given the unfortunate proliferation of branches in
never knows which branch your users will be reporting the bugs in.
That said, you're right -- I should have been more explicit and push for it
to be included in .205 (honestly, I'm really surprised this is like the 4th time
Cos has to commit the same fix to a different branch :-().
> Testing a release candidate is a community effort,
> and out of all the people in the community who tested the RC, no one found
> or raised this as an issue.
I find this really difficult to believe. Either I'm doing something incorrectly,
or you guys had 0 testing with unsecured deployments of Hadoop.
If there's a way to start an unsecure datanode without running into this
issue I'd appreciate it very much if you could tell me how.
>.And yes, I am considering a 0.20.205.1 bug fix release, so if a tested patch
> is submitted in the near future, it can go in.
I would appreciate that. What's the time frame for the release?
> Regarding the second issue, "unfortunate placement of the task-controller
> binary", again please open a jira and submit a tested patch. I'll be happy
> to include it in the next release.
> Thank you for raising these issues. I hope in the future the Bigtop team
> will consider testing the next Hadoop release candidate before the vote
> closes instead of after.
That is, actually, predicated on keeping us in the loop as part of the release
criteria. If you feel that Bigtop adds value to the hadoop release process
I'd love to collaborate with you on any future releases that you might
want to do.
Does that make sense?