looking for help on HBASE

[Sean Busbey <bu...@cloudera.com>]

Hi Phoenix!

I've been trying to chase down the root cause of an issue that y'all
reported with HBase increments that have custom time ranges in
1.2+[1]. Right now this issue is marked as a Blocker and HBase is
waiting on it to continue our 1.2.z releases and start our 1.3.z
release line.

Long story short, thus far my attempts to come up with a unit test
that shows the issue all pass on the HBase side, though I can clearly
see the problem in one of Phoenix's ITs. I've been trying to track
what happens across HBase + Phoenix along the write path but so far I
haven't found a smoking gun.

Could someone familiar with the Phoenix customizations along the HBase
write path spare a bit of time to walk me through things? I want to
make sure I'm looking at the correct places on the Phoenix side.

[1]: https://issues.apache.org/jira/browse/HBASE-15698

-- 
busbey

Re: looking for help on HBASE

[Sergey Soldatov <se...@gmail.com>]

Hey Sean,
I'll take a look while other guys are on HBaseCon.

Thanks,
Sergey

On Tue, May 24, 2016 at 4:49 PM, Sean Busbey <bu...@cloudera.com> wrote:

> Hi Phoenix!
>
> I've been trying to chase down the root cause of an issue that y'all
> reported with HBase increments that have custom time ranges in
> 1.2+[1]. Right now this issue is marked as a Blocker and HBase is
> waiting on it to continue our 1.2.z releases and start our 1.3.z
> release line.
>
> Long story short, thus far my attempts to come up with a unit test
> that shows the issue all pass on the HBase side, though I can clearly
> see the problem in one of Phoenix's ITs. I've been trying to track
> what happens across HBase + Phoenix along the write path but so far I
> haven't found a smoking gun.
>
> Could someone familiar with the Phoenix customizations along the HBase
> write path spare a bit of time to walk me through things? I want to
> make sure I'm looking at the correct places on the Phoenix side.
>
> [1]: https://issues.apache.org/jira/browse/HBASE-15698
>
> --
> busbey
>

Re: looking for help on HBASE

[Sergey Soldatov <se...@gmail.com>]

Hi Sean,
Not sure about the unit test, but the fix that cause our issue is
HBASE-15198. Prior it we had connection with cellBlock false, so protobuf
serialized everything, including timeRange. Now cellBlock is true and
buildNoDataRegionAction is used for the serializing of Increament
mutations. And it doesn't even think about timeRange.
If you want to reproduce it with phoenix, just get the workspace from the
git on the commit provided in Jame's comments, build it with -DskipTests,
import into IDE (I would recommend Idea since it's better handle generated
files), place breakpoint in RequestConverter.buildNoDataRegionAction at
 Increment i = (Increment) row;
Now run PhoenixTimeQueryIT  in the debugger. Several steps and you will see
that builder will be started with empty timeRangeBuilder_

Thanks,
Sergey


On Tue, May 24, 2016 at 4:49 PM, Sean Busbey <bu...@cloudera.com> wrote:

> Hi Phoenix!
>
> I've been trying to chase down the root cause of an issue that y'all
> reported with HBase increments that have custom time ranges in
> 1.2+[1]. Right now this issue is marked as a Blocker and HBase is
> waiting on it to continue our 1.2.z releases and start our 1.3.z
> release line.
>
> Long story short, thus far my attempts to come up with a unit test
> that shows the issue all pass on the HBase side, though I can clearly
> see the problem in one of Phoenix's ITs. I've been trying to track
> what happens across HBase + Phoenix along the write path but so far I
> haven't found a smoking gun.
>
> Could someone familiar with the Phoenix customizations along the HBase
> write path spare a bit of time to walk me through things? I want to
> make sure I'm looking at the correct places on the Phoenix side.
>
> [1]: https://issues.apache.org/jira/browse/HBASE-15698
>
> --
> busbey
>