Hi,
Today I tested this routing feature. It did not work as expected.

1. When I execute Group By queries along with WHERE conditions on a plain
cube  built without conditions,  the JDBC driver simply throws an
exception.  I was expecting Kylin to redirect to impala which I had
configured through hive.url configuration parameter. This did not happen. I
see the code that tries olapenumerator and if that catches exception,
redirects to Hive enumerator. But the output I see does not correspond to
the code. What am I missing?

2. For the scenario described above,  I am able to catch the exception in
Java app code and redirect to Impala. This works like a charm and is very
fast. Not sure if self-service BI tools will do all these. Consider
filtering a pivot chart.  This translates to a SQL with a WHERE followed by
a GROUP BY. This is not a rare case but rather commonly occurring case.

3. When I issue 'select * from table',  Kylin returns cube data instead of
giving raw table contents.  This is a bit unexpected.  But I remember
reading somewhere in the forums earlier that this is actually a feature.
Does my memory serve me right?

Thanks,
Best,
Sarnath
Sent from 1+3

On Sep 2, 2016 9:37 AM, "Sarnath K" <[email protected]> wrote:

> Sure. I assume hive routing is enabled by default so that I should be able
> to do 'select *' queries through JDBC. This is what we have been trying
> with 1.1 and t did not work. The webapp does not use routing and it is
> documented. No worries on the webapp part.
>
> Is there any docs on uninstalling Kylin -remove all HBase tables,
> Meta-Data and wipe of all intermediate hive tables etc..?
>
> Also, I am using prebuilt binaries for HBase 0.98.
> We are using HBase 0.98.6 that comes with CDH.. Hope this config works.
> On Sep 2, 2016 6:43 AM, "ShaoFeng Shi" <[email protected]> wrote:
>
>> Hi Sarnath,
>>
>> For kylin.sh on Ubuntu issue, there is an open JIRA (
>> https://issues.apache.org/jira/browse/KYLIN-1978), would you like to
>> contribute your finding as a patch? thank you!
>>
>> The "kylin.route.hive.enabled" isn't in use anymore; what need is just
>> "hive.url", "hive.user" and "hive.password".
>>
>> All the config parameters can be found in the java
>> class org.apache.kylin.common.KylinConfigBase.java; There is a work item
>> about adding doc for these parameters, we will add it soon.
>>
>> Changing kylin.properties need restart kylin server to take effective;
>> Besides I havn't heard such problem (even with 1.1), please retry and
>> report problem with detail information and we can have a check.
>>
>>
>> 2016-09-01 23:18 GMT+08:00 Sarnath K <[email protected]>:
>>
>> > I just downloaded 1.5.3 version of Kylin. The startup scripts fail in
>> > Ubuntu because of "sh" in place of "bash". Replaced that in 2 scripts to
>> > get going.
>> >
>> > And, apart from hive.URL and other variables, should I also set
>> > kylin.route.hive.enabled to true to enable routing to Hive?
>> >
>> > Where can I find a list of all config vars? Looks like the
>> documentation is
>> > very sparse.
>> >
>> > Also, the webapp does not reflect kylin.properties well. Even after
>> > refreshes.... At least this was the case with 1.1incubating.
>> > On Aug 30, 2016 12:40 PM, "ShaoFeng Shi" <[email protected]>
>> wrote:
>> >
>> > > Hi Sarnath, I totally agree with you; we're looking forward to see
>> your
>> > > sharing and patch.
>> > >
>> > > 2016-08-30 13:47 GMT+08:00 Sarnath K <[email protected]>:
>> > >
>> > > > Sure. I just hope this never goes away. It is a great feature and BI
>> > > tools
>> > > > like Tableau can work seamlessly. And, it is a point I keep telling
>> > > > customers too. It is a great optimization that is really awesome.
>> > > > I will test it with Impala and post back.
>> > > >
>> > >
>> > >
>> > >
>> > > --
>> > > Best regards,
>> > >
>> > > Shaofeng Shi
>> > >
>> >
>>
>>
>>
>> --
>> Best regards,
>>
>> Shaofeng Shi
>>
>

Reply via email to