impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dimitris Tsirogiannis (Code Review)" <>
Subject [Impala-ASF-CR] IMPALA-3739: Enable stress tests on Kudu
Date Thu, 15 Sep 2016 03:33:10 GMT
Dimitris Tsirogiannis has posted comments on this change.

Change subject: IMPALA-3739: Enable stress tests on Kudu

Patch Set 3:

File testdata/bin/

PS3, Line 51:     tbls_to_clean = tpch_tables if workload.lower() == 'tpch' else tpcds_tables
> Maybe use the cursor to get the list of tables? That way you don't have to 
The change to enable drop db cascade for Kudu is in review. If you're strongly against simply
doing a drop db when this is in, I'll implement your proposal.

PS3, Line 81:   sql_file_path = "%s/testdata/datasets/%s/%s_kudu_template.sql"
> Use os.path.join() here.
File testdata/datasets/tpcds/tpcds_kudu_template.sql:

PS3, Line 39: 'kudu.key_columns' = 'ss_sold_date_sk,ss_ticket_number, ss_item_sk'
> For my education, I looked at
No intention here, I just got this wrong :) I actually found a few other inconsistencies.
File testdata/workloads/tpcds/queries/tpcds-kudu-q19.test:

Line 39: ====
> I noticed none of the TPC-DS Kudu queries have RESULTS. Why? (I searched fo
The TPC-DS workload is not currently enabled for Kudu in our regular (non-stress) testing
framework. It requires some extra work which I left for another patch. Once this is done,
these queries will also get proper RESULTS and TYPES sections. For now it seemed kind of pointless
to add these sections. Besides the stress test has its own result verification mechanism.
File testdata/workloads/tpcds/queries/tpcds-kudu-q47.test:

PS3, Line 33: 	,round(v1_lead.sum_sales, 2) nsum
> Nit: tab character.
File testdata/workloads/tpcds/queries/tpcds-kudu-q65.test:

PS3, Line 55: order by
            :   s_store_name,
            :   i_item_desc,
            :   sc.revenue,
            :   i_current_price,
            :   i_wholesale_cost,
            :   i_brand
> The ORDER BY has more columns than the TPC-DS-for-HDFS counterpart. Any rea
I had trouble getting consistent results across multiple invocations of this query unless
I included these additional columns here. Added a comment.
File tests/stress/

PS3, Line 1463:     tpch_kudu_queries = load_tpc_queries("tpch", "kudu")
> Change "kudu" to load_in_kudu=True
Good catch, thanks. Done

PS3, Line 1468:     tpcds_kudu_queries = load_tpc_queries("tpcds", "kudu")
> Change "kudu" to load_in_kudu=True

To view, visit
To unsubscribe, visit

Gerrit-MessageType: comment
Gerrit-Change-Id: I3c9fc3dae24b761f031ee8e014bd611a49029d34
Gerrit-PatchSet: 3
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Dimitris Tsirogiannis <>
Gerrit-Reviewer: Dimitris Tsirogiannis <>
Gerrit-Reviewer: Matthew Jacobs <>
Gerrit-Reviewer: Michael Brown <>
Gerrit-HasComments: Yes

View raw message