[ 
https://issues.apache.org/jira/browse/HBASE-11730?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14095685#comment-14095685
 ] 

Andrew Purtell commented on HBASE-11730:
----------------------------------------

bq. From what I have seen on the jira, it looks like each branch's release 
manager makes the call on backporting a particular issue.

Yes. My understanding of Apache foundation level process is the individuals in 
the RM role assemble release candidate artifacts for voting on by the PMC. What 
the RM puts together can be anything that can be traced back to a revision in 
SCM. This will only work if the RM and PMC are in general agreement, so the RM 
has in effect a filtering function delegated by the PMC. In HBase we don't have 
anything formalized beyond the basics, which I think is a good thing. Whether 
the RM wishes to assert that filtering function prior to commit to a branch or 
after is up to the individual. Our general practice is to ping a RM and wait 
for acknowledgement before backport of discretionary changes. Important bug 
fixes don't need this.

My personal preference is I'd like an opportunity to review significant changes 
before commit but in all other cases please feel free to commit to 0.98 branch, 
I'll handle changes on a C-T-R basis.


> Document release managers for non-deprecated branches
> -----------------------------------------------------
>
>                 Key: HBASE-11730
>                 URL: https://issues.apache.org/jira/browse/HBASE-11730
>             Project: HBase
>          Issue Type: Task
>          Components: documentation
>            Reporter: Sean Busbey
>
> New development goes against trunk and is backported as desired to existing 
> release branches. From what I have seen on the jira, it looks like each 
> branch's release manager makes the call on backporting a particular issue.
> We should document both this norm and who the relevant release manager is for 
> each branch.
> In the current docs, I'd suggest adding the RM list to the "Codelines" 
> section (18.11.1) and add a brief explanation of pinging the RM as a new 
> section after "submitting a patch again" (18.12.6).
> Post HBASE-4593, the note about pinging a prior branch RM should just go as a 
> bullet in the "patch workflow."



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to