Arun,
Please run the command Ankit mentioned in an HBase shell and post the
output back here.
Thanks,
James
On Wednesday, April 20, 2016, Arun Kumaran Sabtharishi <arun1...@gmail.com>
wrote:
> !describe SYSTEM.CATALOG is not returning IS_ROW_TIMESTAMP column.
>
> But we do see this column from select statement:
>
> select * from SYSTEM.CATALOG where TABLE_NAME=’TEST_TABLE_1’ AND
> TABLE_SCHEM IS NULL AND TENANT_ID IS NULL ;
>
>
>
> Thanks,
>
> Arun
>
> On Wed, Apr 20, 2016 at 1:37 AM, Ankit Singhal <ankitsingha...@gmail.com
> <javascript:_e(%7B%7D,'cvml','ankitsingha...@gmail.com');>> wrote:
>
>> Hi Arun,
>>
>> Do you see 'IS_ROW_TIMESTAMP' column in SYSTEM.CATALOG, by doing
>> !describe on system.catalog.
>>
>>
>> if not,
>> can you share the output of below command. As it seems SYSTEM.CATALOG was
>> updated with timestamp greater v4.6 timestamp , and which stopping upgrade
>> code to add a new column.
>>
>> scan 'SYSTEM.CATALOG', {RAW=>true}
>>
>>
>>
>> Regards,
>> Ankit Singhal
>>
>> On Wed, Apr 20, 2016 at 4:25 AM, Arun Kumaran Sabtharishi <
>> arun1...@gmail.com <javascript:_e(%7B%7D,'cvml','arun1...@gmail.com');>>
>> wrote:
>>
>>> After further investigation, we found that Phoenix Upsert query
>>> SYSTEM.CATALOG has IS_ROW_TIMESTAMP column, but PTableImpl.getColumn() is
>>> failing with error:"Undefined column. columnName=IS_ROW_TIMESTAMP" . Does
>>> this mean that PTableImpl is reading from cached entity of SYSTEM.CATALOG
>>> before 4.6 upgrade?"
>>>
>>> We do see that clearCache() is being called for 4.7, and 4.7 upgrades
>>> from ConnectionQueryServicesImpl class, but not for 4.6
>>>
>>>
>>> Thanks,
>>> Arun
>>>
>>> On Tue, Apr 19, 2016 at 10:22 AM, Arun Kumaran Sabtharishi <
>>> arun1...@gmail.com <javascript:_e(%7B%7D,'cvml','arun1...@gmail.com');>>
>>> wrote:
>>>
>>>> James,
>>>>
>>>> To add more information on this issue, this happens in new phoenix
>>>> views associated with brand new tables as well. So, this cannot be an
>>>> upgrade/migration issue. Not figured out a specific way to reproduce this
>>>> issue yet. Could you throw some ideas on what direction this problem could
>>>> be approached from this point?
>>>>
>>>> Thanks,
>>>> Arun
>>>>
>>>
>>>
>>
>
