hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stack <st...@duboce.net>
Subject Re: [DISCUSS] Best way to help out with 2.0?
Date Thu, 07 Sep 2017 04:35:39 GMT
On Wed, Sep 6, 2017 at 8:18 PM, Josh Elser <elserj@apache.org> wrote:

> Hi folks,
>
> I find myself in a position where I have a bit more time to start
> investing into HBase 2.0, but I find myself a bit stymied about where
> to begin. I appreciate the document tracking the high-level features
> that Stack has been maintaining (and reminding the rest of us to
> update), but it's a bit too high-level when I'm looking for something
> concrete to debug/fix and advance the progress.
>
> I can imagine that there are other folks in the same scenario, so I
> figured it would be good to ask the question to the group at large.
> Some initial thoughts I had on the matter...
>
> * Is a JIRA search for fixVersion=2.0, unassigned, and unresolved
> meaningful?
> * Is looking at busted tests on Jenkins worthwhile? I'd worry about
> working on something someone else is already debugging.
>
>

Thank you for asking Josh.

Some notes for context (some repeated info from other, concurrent thread
but I figure that it is ok given thread title):

+ Plan is to do three alphas (two of which have already released) and a few
betas before we put up an RC (Alphas and betas have the usual connotations
-- shout if you need exposition)
+ Up in JIRA, there is the 2.0.0 fix version. A few passes have pared down
the 2.0.0 set. My guess is that ultimately 1/4 of the JIRAs will make it.
We are engineers after all.
+ We've also been keeping versions in JIRA for each particular step on the
way to 2.0.0 so you'll see a 2.0.0-alpha-1, -alpha-2, and so on.
+ We are currently trying to get 2.0.0-alpha-3 out. The theme for this
release is API stabilization. The plan is to put up an -alpha-3 over the
weekend.
+ Here is the state of -alpha-3 [1]. The 'issues in progress' are doing
well EXCEPT for the monster HBASE-18298
<https://issues.apache.org/jira/browse/HBASE-18298>, the Coprocessor APIs
are a total mess issue, and need severe cleanup (Anoop and Duo have been
doing a good job plugging away at this so far but needs more shoulders
applied).
+ Of the 22 'Issues to do', near all are unassigned.
+ If you are looking for high-payback, picking up one of these issues will
earn you boocoo [2] 'RM dollars' [3].

Otherwise, pick anything that grabs your fancy out of the 2.0.0 bag.

Busted tests on Jenkins will always double your (or anyone else's)
karma.... Quadruple your karma by pointing at a test that overlaps doing
same thing and removing it or by cutting test time by some discernible
percentage.

The "Book of hbase-2.0.0" goes on [4]. It now has fat section on the
incompatibilities between 1.x and 2.x that is in need of tidying. The short
version is that Admin and Coprocessors are irretrievably incompatible but
hbase1 clients should be able to do DML against an hbase2 cluster and
rolling upgrade while DML still seems possible (We'll see). The final
chapter, Migration (hopefully Auto-Migration) is yet to be written.

Hope this helps. Ask if you need me to natter more.

S


1. https://issues.apache.org/jira/projects/HBASE/versions/12340860
2.  "...the illiterate Americanized version of the French adverb "beaucoup"
meaning much or many, that was used in Vietnam."
http://www.urbandictionary.com/
3. New currency I'm about to launch to fund a startup that will make me
rich outsourcing release management to open source projects.
4.
https://docs.google.com/document/d/1WCsVlnHjJeKUcl7wHwqb4z9iEu_ktczrlKHK8N4SZzs/edit#heading=h.7oxmbxtjy4s2

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message