Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/18676 )

Change subject: IMPALA-11430: Support custom hash schema for Kudu range tables
......................................................................


Patch Set 11:

(3 comments)

http://gerrit.cloudera.org:8080/#/c/18676/11/testdata/workloads/functional-query/queries/QueryTest/kudu_partition_ddl.test
File 
testdata/workloads/functional-query/queries/QueryTest/kudu_partition_ddl.test:

http://gerrit.cloudera.org:8080/#/c/18676/11/testdata/workloads/functional-query/queries/QueryTest/kudu_partition_ddl.test@377
PS11, Line 377: '30 <= VALUES < 40 HASH(id,c2) PARTITIONS 2'
> The hash parameters at both the table and partition level default to the PK
Ah, great: that makes sense.


http://gerrit.cloudera.org:8080/#/c/18676/11/testdata/workloads/functional-query/queries/QueryTest/kudu_partition_ddl.test@380
PS11, Line 380: ====
> added testcases. The PARTITIONS 0 and PARTITIONS 1 cases were flakey, somet
It should always return an error -- there isn't any magic in that part and I 
don't see how it can be flakey :)  I guess the root cause was that the Kudu 
binaries in your test cluster were missing this changelist: 
https://gerrit.cloudera.org/#/c/18728/

Without the patch, an attempt to create a table with incorrect number of hash 
bucket in table-wide hash schema would return an error, but a request to add a 
range partition with custom hash schema with incorrect number of hash buckets 
would be accepted, but IsAlterTable done would time out.

By the time you were working on this, the patch hadn't yet been committed into 
the Kudu upstream git repo.

As discussed offline, I added testcases into the Kudu codebase to cover that 
functionality, and everything works as expected: 
https://gerrit.cloudera.org/#/c/18739/

As of now, https://gerrit.cloudera.org/#/c/18728/ is already committed into the 
Kudu upstream repo, so these negative tests cases can now be added.


http://gerrit.cloudera.org:8080/#/c/18676/17/testdata/workloads/functional-query/queries/QueryTest/kudu_partition_ddl.test
File 
testdata/workloads/functional-query/queries/QueryTest/kudu_partition_ddl.test:

http://gerrit.cloudera.org:8080/#/c/18676/17/testdata/workloads/functional-query/queries/QueryTest/kudu_partition_ddl.test@322
PS17, Line 322: ---- QUERY
Does it make sense to add a test case to add a range into a table with 
non-empty table-wide hash schema (similar to what's created at lines 298-305) 
without information on hash parameters?  The query has only range-related 
parameters:

  ALTER TABLE custom_hash_range ADD RANGE PARTITION 300 <= VALUES < 400;

Correct me if I got it wrong, but there should be an error since the DDL query 
would try to add a range with empty hash schema (i.e. zero hash dimensions), 
while the table-wide hash schema has two hash dimensions, and as of now Kudu 
doesn't support varying number of hash dimensions for hash schemas of the same 
table.



--
To view, visit http://gerrit.cloudera.org:8080/18676
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I981056e0827f4957580706d6e73742e4e6743c1c
Gerrit-Change-Number: 18676
Gerrit-PatchSet: 11
Gerrit-Owner: Kurt Deschler <[email protected]>
Gerrit-Reviewer: Abhishek Chennaka <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Joe McDonnell <[email protected]>
Gerrit-Reviewer: Kurt Deschler <[email protected]>
Gerrit-Reviewer: Mahesh Reddy <[email protected]>
Gerrit-Reviewer: Michael Smith <[email protected]>
Gerrit-Reviewer: Quanlong Huang <[email protected]>
Gerrit-Comment-Date: Sat, 16 Jul 2022 01:51:26 +0000
Gerrit-HasComments: Yes

Reply via email to