Hi Devs

After the first RC for 0.90.6 was taken
HBASE-5196 - Failure in region split after PONR could cause region hole
HBASE-5153 - Add retry logic in HConnectionImplementation#resetZooKeeperTrackers

The above 2 defects have been committed.

HBASE-5179 - Concurrent processing of processFaileOver and ServerShutdownHandler may cause region to be assigned before log splitting is completed, causing data loss
HBASE-5225 - Backport HBASE-3845 -data loss because lastSeqWritten can miss memstore edits
HBASE-5235 - HLogSplitter writer thread's streams not getting closed when any of the writer threads has exceptions.
HBASE-5237 - Addendum for HBASE-5160 and HBASE-4397

HBASE-5179 - is almost in a final stage for committing.  Thanks to Chunhui, Ted and Jinchao for persisting on the defect.

The above defects were found during the testing for RC0. Hence i would like to cut another RC once the above
defects goes into 0.90.  By tomorrow 22nd January i would like to take a release cut.
Please let me know your suggestions/opinions.

Regards
Ram

Stack said he would come up with some test for hbase-5179.
Suppose that takes a few more days, do you plan to check in the fix into 0.90 branch ?

According to Todd's suggestion earlier, a Jira shouldn't be open for too long during which time patches continuously get checked in.

Cheers

On Jan 21, 2012, at 4:34 AM, Ramakrishna s vasudevan <[EMAIL PROTECTED]> wrote:

> Hi Devs
>
> After the first RC for 0.90.6 was taken
> HBASE-5196 - Failure in region split after PONR could cause region hole
> HBASE-5153 - Add retry logic in HConnectionImplementation#resetZooKeeperTrackers
>
> The above 2 defects have been committed.
>
> HBASE-5179 - Concurrent processing of processFaileOver and ServerShutdownHandler may cause region to be assigned before log splitting is completed, causing data loss
> HBASE-5225 - Backport HBASE-3845 -data loss because lastSeqWritten can miss memstore edits
> HBASE-5235 - HLogSplitter writer thread's streams not getting closed when any of the writer threads has exceptions.
> HBASE-5237 - Addendum for HBASE-5160 and HBASE-4397
>
> HBASE-5179 - is almost in a final stage for committing.  Thanks to Chunhui, Ted and Jinchao for persisting on the defect.
>
> The above defects were found during the testing for RC0. Hence i would like to cut another RC once the above
> defects goes into 0.90.  By tomorrow 22nd January i would like to take a release cut.
> Please let me know your suggestions/opinions.
>
> Regards
> Ram

It is always better to get in a patch with test case.  But if it takes  a little more time to get the test case can we verify the patch with good cluster testing and raise a JIRA for the test case integration that Stack gives.

By this way we can get the patch in the release and also satisfies Todd's suggestion.

Any comments so that i can raise a test task for the same.

Regards
Ram
________________________________________
From: [EMAIL PROTECTED] [[EMAIL PROTECTED]]
Sent: Saturday, January 21, 2012 8:13 PM
To: [EMAIL PROTECTED]
Cc: [EMAIL PROTECTED]
Subject: Re: 0.90.6 Release status

Stack said he would come up with some test for hbase-5179.
Suppose that takes a few more days, do you plan to check in the fix into 0.90 branch ?

According to Todd's suggestion earlier, a Jira shouldn't be open for too long during which time patches continuously get checked in.

Cheers

On Jan 21, 2012, at 4:34 AM, Ramakrishna s vasudevan <[EMAIL PROTECTED]> wrote:

> Hi Devs
>
> After the first RC for 0.90.6 was taken
> HBASE-5196 - Failure in region split after PONR could cause region hole
> HBASE-5153 - Add retry logic in HConnectionImplementation#resetZooKeeperTrackers
>
> The above 2 defects have been committed.
>
> HBASE-5179 - Concurrent processing of processFaileOver and ServerShutdownHandler may cause region to be assigned before log splitting is completed, causing data loss
> HBASE-5225 - Backport HBASE-3845 -data loss because lastSeqWritten can miss memstore edits
> HBASE-5235 - HLogSplitter writer thread's streams not getting closed when any of the writer threads has exceptions.
> HBASE-5237 - Addendum for HBASE-5160 and HBASE-4397
>
> HBASE-5179 - is almost in a final stage for committing.  Thanks to Chunhui, Ted and Jinchao for persisting on the defect.
>
> The above defects were found during the testing for RC0. Hence i would like to cut another RC once the above
> defects goes into 0.90.  By tomorrow 22nd January i would like to take a release cut.
> Please let me know your suggestions/opinions.
>
> Regards
> Ram

(Coming in a little late, sorry about that)

If HBASE-5196 is a regression from previous versions, then definitely
-1. If not, I don't see why we'd sink the RC and even more so since it
has no unit test which means it could add more bugs to a bug fix
release.

That said, you guys can still -1 this RC if you want.

Personally I'm done with the 0.90 branch and I'm focussing my efforts
on getting 0.92 in production here.

J-D

On Sat, Jan 21, 2012 at 8:38 AM, Ramakrishna s vasudevan
<[EMAIL PROTECTED]> wrote:
> It is always better to get in a patch with test case.  But if it takes  a little more time to get the test case can we verify the patch with good cluster testing and raise a JIRA for the test case integration that Stack gives.
>
> By this way we can get the patch in the release and also satisfies Todd's suggestion.
>
> Any comments so that i can raise a test task for the same.
>
> Regards
> Ram
> ________________________________________
> From: [EMAIL PROTECTED] [[EMAIL PROTECTED]]
> Sent: Saturday, January 21, 2012 8:13 PM
> To: [EMAIL PROTECTED]
> Cc: [EMAIL PROTECTED]
> Subject: Re: 0.90.6 Release status
>
> Stack said he would come up with some test for hbase-5179.
> Suppose that takes a few more days, do you plan to check in the fix into 0.90 branch ?
>
> According to Todd's suggestion earlier, a Jira shouldn't be open for too long during which time patches continuously get checked in.
>
> Cheers
>
>
>
> On Jan 21, 2012, at 4:34 AM, Ramakrishna s vasudevan <[EMAIL PROTECTED]> wrote:
>
>> Hi Devs
>>
>> After the first RC for 0.90.6 was taken
>> HBASE-5196 - Failure in region split after PONR could cause region hole
>> HBASE-5153 - Add retry logic in HConnectionImplementation#resetZooKeeperTrackers
>>
>> The above 2 defects have been committed.
>>
>> HBASE-5179 - Concurrent processing of processFaileOver and ServerShutdownHandler may cause region to be assigned before log splitting is completed, causing data loss
>> HBASE-5225 - Backport HBASE-3845 -data loss because lastSeqWritten can miss memstore edits
>> HBASE-5235 - HLogSplitter writer thread's streams not getting closed when any of the writer threads has exceptions.
>> HBASE-5237 - Addendum for HBASE-5160 and HBASE-4397
>>
>> HBASE-5179 - is almost in a final stage for committing.  Thanks to Chunhui, Ted and Jinchao for persisting on the defect.
>>
>> The above defects were found during the testing for RC0. Hence i would like to cut another RC once the above
>> defects goes into 0.90.  By tomorrow 22nd January i would like to take a release cut.
>> Please let me know your suggestions/opinions.
>>
>> Regards
>> Ram

Hey Ram,

I'm still interested in the 0.90 branch and will spend some time testing
the RC.

Keep up the good work!

Jon.

On Tue, Jan 24, 2012 at 9:57 AM, Jean-Daniel Cryans <[EMAIL PROTECTED]>wrote:

> (Coming in a little late, sorry about that)
>
> If HBASE-5196 is a regression from previous versions, then definitely
> -1. If not, I don't see why we'd sink the RC and even more so since it
> has no unit test which means it could add more bugs to a bug fix
> release.
>
> That said, you guys can still -1 this RC if you want.
>
> Personally I'm done with the 0.90 branch and I'm focussing my efforts
> on getting 0.92 in production here.
>
> J-D
>
> On Sat, Jan 21, 2012 at 8:38 AM, Ramakrishna s vasudevan
> <[EMAIL PROTECTED]> wrote:
> > It is always better to get in a patch with test case.  But if it takes
>  a little more time to get the test case can we verify the patch with good
> cluster testing and raise a JIRA for the test case integration that Stack
> gives.
> >
> > By this way we can get the patch in the release and also satisfies
> Todd's suggestion.
> >
> > Any comments so that i can raise a test task for the same.
> >
> > Regards
> > Ram
> > ________________________________________
> > From: [EMAIL PROTECTED] [[EMAIL PROTECTED]]
> > Sent: Saturday, January 21, 2012 8:13 PM
> > To: [EMAIL PROTECTED]
> > Cc: [EMAIL PROTECTED]
> > Subject: Re: 0.90.6 Release status
> >
> > Stack said he would come up with some test for hbase-5179.
> > Suppose that takes a few more days, do you plan to check in the fix into
> 0.90 branch ?
> >
> > According to Todd's suggestion earlier, a Jira shouldn't be open for too
> long during which time patches continuously get checked in.
> >
> > Cheers
> >
> >
> >
> > On Jan 21, 2012, at 4:34 AM, Ramakrishna s vasudevan <
> [EMAIL PROTECTED]> wrote:
> >
> >> Hi Devs
> >>
> >> After the first RC for 0.90.6 was taken
> >> HBASE-5196 - Failure in region split after PONR could cause region hole
> >> HBASE-5153 - Add retry logic in
> HConnectionImplementation#resetZooKeeperTrackers
> >>
> >> The above 2 defects have been committed.
> >>
> >> HBASE-5179 - Concurrent processing of processFaileOver and
> ServerShutdownHandler may cause region to be assigned before log splitting
> is completed, causing data loss
> >> HBASE-5225 - Backport HBASE-3845 -data loss because lastSeqWritten can
> miss memstore edits
> >> HBASE-5235 - HLogSplitter writer thread's streams not getting closed
> when any of the writer threads has exceptions.
> >> HBASE-5237 - Addendum for HBASE-5160 and HBASE-4397
> >>
> >> HBASE-5179 - is almost in a final stage for committing.  Thanks to
> Chunhui, Ted and Jinchao for persisting on the defect.
> >>
> >> The above defects were found during the testing for RC0. Hence i would
> like to cut another RC once the above
> >> defects goes into 0.90.  By tomorrow 22nd January i would like to take
> a release cut.
> >> Please let me know your suggestions/opinions.
> >>
> >> Regards
> >> Ram
>

--
// Jonathan Hsieh (shay)
// Software Engineer, Cloudera
// [EMAIL PROTECTED]