hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Yu <yuzhih...@gmail.com>
Subject Re: [VOTE] HBase 1.2.0 RC2
Date Fri, 12 Feb 2016 21:28:15 GMT
I did some validation today.
See my comment at the tail of HBASE-15219.

I reverted my patch until we find out which patch caused the
-treatFailureAsError
flag not to work.

On Fri, Feb 12, 2016 at 1:22 PM, Stack <stack@duboce.net> wrote:

> I just ran a small ITBLL against current 1.2 HEAD and it seems fine...
> nothing untoward in logs. Running bigger one now. Lets just go w/ tip of
> 1.2? And one of the items just got reverted:
>
> commit e52ac92b9810425cb5345121260959e4c0ad5ab3
> Author: tedyu <yuzhihong@gmail.com>
> Date:   Fri Feb 12 12:01:45 2016 -0800
>
>     HBASE-15219 Revert pending verification of test result
>
> St.Ack
>
>
> On Fri, Feb 12, 2016 at 11:44 AM, Sean Busbey <busbey@cloudera.com> wrote:
>
> > here is what has happened on branch-1.2 since RC2:
> >
> > * 7ed1603 - (origin/branch-1.2) HBASE-15252 Data loss when replaying wal
> if
> > HDFS timeout (11 hours ago)
> > * 19d964d - HBASE-15198 RPC client not using Codec and CellBlock for puts
> > by default-addendum. (18 hours ago)
> > * cc863f3 - HBASE-15224 Undo
> "hbase.increment.fast.but.narrow.consistency"
> > option; it is not necessary since HBASE-15213 (23 hours ago)
> > * 644326b - HBASE-15129 Set default value for hbase.fs.tmp.dir rather
> than
> > fully depend on hbase-default.xml (Yu Li) (27 hours ago)
> > * 7d5a158 - HBASE-15198 RPC client not using Codec and CellBlock for puts
> > by default. (33 hours ago)
> > * c5b6c96 - HBASE-14192 Fix REST Cluster Constructor with String List (2
> > days ago)
> > * 3b6c305 - HBASE-15229 Canary Tools should not call System.Exit on error
> > (Vishal Khandelwal) (2 days ago)
> > * 8a2cb16 - HBASE-15219 Canary tool does not return non-zero exit code
> when
> > one of regions is in stuck state (2 days ago)
> > * 7643509 - HBASE-15216 Canary does not accept config params from command
> > line (Vishal Khandelwal) (3 days ago)
> > * d5fd993 - HBASE-15238 HFileReaderV2 prefetch overreaches; runs off the
> > end of the data; ADDENDUM (3 days ago)
> > * 6f6cd66 -     HBASE-15238 HFileReaderV2 prefetch overreaches; runs off
> > the end of the data (3 days ago)
> > * 4cb21cf - HBASE-15224 Undo
> "hbase.increment.fast.but.narrow.consistency"
> > option; it is not necessary since HBASE-15213 (4 days ago)
> > * d568db8 - (1.2.0RC2) HBASE-14025 update CHANGES.txt for 1.2 RC2 (5 days
> > ago)
> >
> > I *could* make 1.2.0 RC3 that just cherry picks HBASE-15252 onto RC2, but
> > that's going to make things a bit messy and possibly confusing for folks
> > who look for the 1.2.0 tag to be an ancestor of branch-1.2's HEAD.
> >
> > On Fri, Feb 12, 2016 at 1:16 PM, Andrew Purtell <
> andrew.purtell@gmail.com>
> > wrote:
> >
> > > Same here. I have started with RC2 but can mostly carry findings to RC3
> > > given only one additional change.
> > >
> > > > On Feb 12, 2016, at 8:56 AM, Elliott Clark <eclark@apache.org>
> wrote:
> > > >
> > > > -1 until the dataloss is fixed.
> > > >
> > > > But assuming that's fixed I would be good for a short vote cycle for
> > the
> > > > next RC.
> > > >
> > > >> On Fri, Feb 12, 2016 at 1:02 AM, 张铎 <palomino219@gmail.com>
wrote:
> > > >>
> > > >> HBASE-15252 is fixed :).
> > > >>
> > > >> 2016-02-12 14:00 GMT+08:00 Stack <stack@duboce.net>:
> > > >>
> > > >>> -1
> > > >>>
> > > >>> The dataloss issue was just discovered. I think now we know of
it,
> > even
> > > >>> though the incidence is rare, would be best to respin the RC.
> > > >>>
> > > >>> You the man Sean,
> > > >>> St.Ack
> > > >>>
> > > >>>
> > > >>>> On Thu, Feb 11, 2016 at 8:59 PM, Stack <stack@duboce.net>
wrote:
> > > >>>>
> > > >>>> On Thu, Feb 11, 2016 at 5:04 PM, Sean Busbey <
> sean.busbey@gmail.com
> > >
> > > >>>> wrote:
> > > >>>>
> > > >>>>>> On Feb 11, 2016 18:33, "张铎" <palomino219@gmail.com>
wrote:
> > > >>>>>>
> > > >>>>>> Should we include HBASE-15252? It is a data loss issue.
> > > >>>>>
> > > >>>>> It's marked major (though perhaps that's off since it's
dataloss,
> > > even
> > > >>> if
> > > >>>>> rare). More importantly it's been present in prior releases
for
> > some
> > > >>> time.
> > > >>>>>
> > > >>>>> Blocking 1.2.0 would put pressure on getting a solution
faster, I
> > > >> think.
> > > >>>>> Additionally, letting the fix wait for 1.2.1 will give
me a good
> > > >>> incentive
> > > >>>>> to keep the path releases on schedule. ;)
> > > >>>>>
> > > >>>>> My 2¢. Happy to roll another RC if folks see it otherwise.
> > > >>>>
> > > >>>> Dataloss. I think we should roll a new RC with short voting
> > timeframe.
> > > >>>> St.Ack
> > > >>
> > >
> >
> >
> >
> > --
> > Sean
> >
>

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