[ 
https://issues.apache.org/jira/browse/IMPALA-1995?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17540718#comment-17540718
 ] 

Quanlong Huang commented on IMPALA-1995:
----------------------------------------

Take the "functional_hbase.alltypesagg" table as an example, we can first list 
all the regions of it. The output is ordered by the start key. Then we just 
need to move every 2 regions to one RegionServer.

{code}
hbase:001:0> list_regions "functional_hbase.alltypesagg"
                   SERVER_NAME |                                                
                    REGION_NAME |  START_KEY |    END_KEY |  SIZE |    REQ |   
LOCALITY |
 ----------------------------- | 
------------------------------------------------------------------------------ 
| ---------- | ---------- | ----- | ------ | ---------- |
 localhost,16020,1653261458193 |  
functional_hbase.alltypesagg,,1653265020292.89f28ad00353462786ce54d89553c1ff. | 
           |          1 |     1 |     90 |        1.0 |
 localhost,16020,1653261458193 | 
functional_hbase.alltypesagg,1,1653265020292.5cc7f766115d323c9bbdf6592e3a2d1c. 
|          1 |          3 |     1 |      0 |        1.0 |
 localhost,16022,1653261459394 | 
functional_hbase.alltypesagg,3,1653265020292.bdd4d0417627790951f61623798b6ec5. 
|          3 |          5 |     1 | 126348 |        1.0 |
 localhost,16022,1653261459394 | 
functional_hbase.alltypesagg,5,1653265020292.43a339608bc7da153a7ef2fe61a5d4ce. 
|          5 |          7 |     1 | 130135 |        1.0 |
 localhost,16023,1653261460504 | 
functional_hbase.alltypesagg,7,1653265020292.46c00822bd24a69709b10fad4055363a. 
|          7 |          9 |     1 |  81468 |        1.0 |
 localhost,16023,1653261460504 | 
functional_hbase.alltypesagg,9,1653265020292.61fd266060bc61909db0440c84b0d703. 
|          9 |            |     1 |  81477 |        1.0 |
 6 rows
Took 0.4263 seconds                                                             
                                                                                
                                                                       
hbase:002:0> move "89f28ad00353462786ce54d89553c1ff", 
"localhost,16020,1653261458193"
Took 0.0214 seconds
hbase:003:0> move "5cc7f766115d323c9bbdf6592e3a2d1c", 
"localhost,16020,1653261458193"
Took 0.0060 seconds                                                             
                                                                                
                                                                       
hbase:004:0> move "bdd4d0417627790951f61623798b6ec5", 
"localhost,16022,1653261459394"
Took 0.0057 seconds                                                             
                                                                                
                                                                       
hbase:005:0> move "43a339608bc7da153a7ef2fe61a5d4ce", 
"localhost,16022,1653261459394"
Took 0.0017 seconds                                                             
                                                                                
                                                                       
hbase:006:0> move "46c00822bd24a69709b10fad4055363a", 
"localhost,16023,1653261460504"
Took 0.0058 seconds                                                             
                                                                                
                                                                       
hbase:007:0> move "61fd266060bc61909db0440c84b0d703", 
"localhost,16023,1653261460504"
Took 0.0022 seconds
{code}

Note that on my desktop, the regions are assigned as expected. So the move 
commands don't change anything. 

> Flaky test: PlannerTest.testHbase: splits for HBASE KEYRANGE not set up 
> correctly.
> ----------------------------------------------------------------------------------
>
>                 Key: IMPALA-1995
>                 URL: https://issues.apache.org/jira/browse/IMPALA-1995
>             Project: IMPALA
>          Issue Type: Task
>          Components: Frontend
>    Affects Versions: Impala 2.3.0, Impala 3.2.0
>            Reporter: Alexander Behm
>            Assignee: Joe McDonnell
>            Priority: Critical
>              Labels: broken-build, flaky, test-infra
>             Fix For: Impala 2.13.0, Impala 3.1.0
>
>
> Looks like testdata/bin/split-hbase.sh does not always set up the ranges in 
> the way we want. See failure below.
> Example run:
> http://sandbox.jenkins.cloudera.com/job/impala-master-cdh5-trunk/1090/
> {code}
> Stack Trace:
> java.lang.AssertionError: section SCANRANGELOCATIONS of query:
> select * from functional_hbase.alltypessmall
> where id < 5
> actual result doesn't match expected result:
>   HBASE KEYRANGE port=16202 3:7
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>   HBASE KEYRANGE port=16203 7:<unbounded>
>   HBASE KEYRANGE port=16203 <unbounded>:3
> NODE 0:
> expected:
>   HBASE KEYRANGE port=16201 <unbounded>:3
>   HBASE KEYRANGE port=16202 3:7
>   HBASE KEYRANGE port=16203 7:<unbounded>
> NODE 0:
> section SCANRANGELOCATIONS of query:
> select * from functional_hbase.stringids
> where id < '5'
> and tinyint_col = 5
> actual result doesn't match expected result:
>   HBASE KEYRANGE port=16201 1:3
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>   HBASE KEYRANGE port=16202 3:5
>   HBASE KEYRANGE port=16203 <unbounded>:1
> NODE 0:
> expected:
>   HBASE KEYRANGE port=16201 <unbounded>:3
>   HBASE KEYRANGE port=16202 3:5
> NODE 0:
> section SCANRANGELOCATIONS of query:
> select * from functional_hbase.alltypesagg
> where bigint_col is not null and bool_col = true
> actual result doesn't match expected result:
>   HBASE KEYRANGE port=16201 1:3
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>   HBASE KEYRANGE port=16202 3:7
>   HBASE KEYRANGE port=16203 7:<unbounded>
>   HBASE KEYRANGE port=16203 <unbounded>:1
> NODE 0:
> expected:
>   HBASE KEYRANGE port=16201 <unbounded>:3
>   HBASE KEYRANGE port=16202 3:7
>   HBASE KEYRANGE port=16203 7:<unbounded>
> NODE 0:
> {code}



--
This message was sent by Atlassian Jira
(v8.20.7#820007)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to