Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase >> mail # dev >> Snapshots branch trunk merging


Copy link to this message
-
Re: Snapshots branch trunk merging
I took a quick look at the megapatch attached to hbase-7290.  Added some
minor remarks to the issue.

+1 on commit as long as hadoopqa is green (caveat the flakies).

St.Ack
On Tue, Feb 12, 2013 at 2:12 PM, Jonathan Hsieh <[EMAIL PROTECTED]> wrote:

> Hey all,
>
> We did a branch merge of the snapshot branch (hbase-6055 and
> hbase-7290 combined) with trunk on 2/1/13.  This merge initially had
> several always broken tests but  since then Ted, Matteo and myself
> fixed all the always-broken unit tests.  I've merged again today
> 2/12/13 [1], and posted a patch on HBASE-7290 for the hadoopqa bot to
> run.
>
> There are primarily four of us who worked on this branch -- Jesse Y,
> Matteo, Ted Yu and myself, so if we each +1, technically we would have
> the 3 +1's required and could merge.  I wanted to solicit +1's from
> the four who worked  on it (all committers now) and also find out if
> anyone else has started reviewing the code or intends to in the next
> few days.  It is a large patch (1.3MB) that I can post on review
> board, but it may be easier to understand by going to the different
> individual jiras (some of which have design docs).   Generally we've
> been using a looser of review then commit for each of the subtasks. If
> I get clean test runs from the QA bot and +1's from the folks who
> worked on it or planned on reviewing it, I'd like merge sooner rather
> than later.
>
> On the unit testing front, I've personally gotten one error-free unit
> tests runs runs and one with a failure in hbase-examples:
>
> org.apache.hadoop.hbase.coprocessor.example.TestBulkDeleteProtocol.testBulkDeleteFamily.
>
> On the system testing side, I've been testing the pre-merge version
> outlined by Aleks [2] and it had been fairly robust on a 5 node
> cluster with fault injection.  I've also done some testing on a 20
> node cluster (no fault injection) and a few runs on a 100 node cluster
> where the snapshoting feature has been robust.
>
> Thanks,
> Jon.
>
> [1] https://github.com/jmhsieh/hbase/tree/snapshot-merge-0212
> [2] http://markmail.org/message/pdbkq654ipuxyt6a
>
> --
> // Jonathan Hsieh (shay)
> // Software Engineer, Cloudera
> // [EMAIL PROTECTED]
>