Re: Phoenix Index tool deletes data from main table and does not populate local index

[Istvan Toth <st...@apache.org>]

Hi,

This list is for the Apache Phoenix releases.
If you use a vendor version, please contact your vendor for support.

regards
Istvan

On Mon, Aug 22, 2022 at 7:21 PM Pradheep Shanmugam via user <
user@phoenix.apache.org> wrote:

> HI,
>
> I am using CDP 7.3.1
> I am trying create a phoenix local index for already existing table with
> ~14M rows.
> Following are the steps I did.
> 1. take snapshot of table from hbase (table is salted in phoenix)
> 2. clone the snapshot to new test table
> 3. create the table in phoenix to link the phoenix to new habse table
> created from clone snapshot. I can see all rows in the phoenix table.
>
>
> 4. create the local index in async mode:
> CREATE LOCAL INDEX MILESTONE_LOCAL_INDEX ON MILESTONE_TEST
> (h.eventTimestamp) ASYNC;
>
>
> 5. Run the MR job Phoenix IndexTool:
>
> ./hbase org.apache.phoenix.mapreduce.index.IndexTool --data-table
> MILESTONE_TEST --index-table MILESTONE_LOCAL_INDEX --output-path
> /hbase/data/default/MILESTONE_LOCAL_INDEX_HFILE
>
> MR jobs says succeeded, but he data in main test table  and local index is
> 0 but the index is marked active.
> I don’t see any explicit error in MR job.
> What could be the issue with MR job?
>
> Another question on Local index usage.
> When I tried with some 5 rows, I tried to see if the local index is used
>
> explain select * from MILESTONE_TEST where eventTimestamp <=
> TO_TIMESTAMP('2022-07-25 14:03:22.559');
>
>
> +-------------------------------------------------------------------------------------------+
>
> |
> PLAN                                            |
>
>
> +-------------------------------------------------------------------------------------------+
>
> | CLIENT 10-CHUNK 0 ROWS 0 BYTES PARALLEL 10-WAY ROUND ROBIN FULL SCAN
> OVER MILESTONE_TEST  |
>
> |     SERVER FILTER BY H.EVENTTIMESTAMP <= TIMESTAMP '2022-07-25
> 14:03:22.559'              |
>
>
> +-------------------------------------------------------------------------------------------+
>
> I am not seeing the local index name in the explain suggesting its usage.
> Does the phoenix not use the local index if there are less rows? Hence am
> trying with millions of rows so that phoenix with try to use the local
> index. Can I expect to see to the local index name in explain if I have a
> few million rows in table and local index?
>
>
> Thanks,
> Pradheep
>