[DRILL-7570] Fix unstable statistics tests

• Type: Task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.17.0
• Fix Version/s: 1.18.0
• Labels: ready-to-commit
• Assignee: Vova Vysotskyi
• Reporter: Vova Vysotskyi
• Reviewer: Arina Ielchiieva
• Created: 05/Feb/20 11:49
• Updated: 06/Feb/20 09:27
• Resolved: 06/Feb/20 09:27

Description

Drill contains tests for checking that statistics is applied, some of them also use sampling to calculate statistics value.

Sampling adds limit above scan, but tests check the value of the estimated row count to verify that statistics were applied. limit without sorting doesn't guarantee consistent results, so these tests may fail sometime:

[ERROR]   TestMetastoreCommands.testAnalyzeWithSampleStatistics:2739 Did not find expected pattern in plan: Filter\(condition.*\).*rowcount = 96.25,
00-00    Screen : rowType = RecordType(ANY employee_id): rowcount = 105.0, cumulative cost = {2530.5 rows, 7570.5 cpu, 2310.0 io, 0.0 network, 0.0 memory}, id = 336738
00-01      Project(employee_id=[$1]) : rowType = RecordType(ANY employee_id): rowcount = 105.0, cumulative cost = {2520.0 rows, 7560.0 cpu, 2310.0 io, 0.0 network, 0.0 memory}, id = 336737
00-02        SelectionVectorRemover : rowType = RecordType(ANY department_id, ANY employee_id): rowcount = 105.0, cumulative cost = {2415.0 rows, 7455.0 cpu, 2310.0 io, 0.0 network, 0.0 memory}, id = 336736
00-03          Filter(condition=[=($0, 2)]) : rowType = RecordType(ANY department_id, ANY employee_id): rowcount = 105.0, cumulative cost = {2310.0 rows, 7350.0 cpu, 2310.0 io, 0.0 network, 0.0 memory}, id = 336735
00-04            Scan(table=[[dfs, tmp, employeeWithStatsFile]], groupscan=[ParquetGroupScan [entries=[ReadEntryWithPath [path=/home/runner/work/drill/drill/exec/java-exec/target/org.apache.drill.exec.sql.TestMetastoreCommands/dfsTestTmp/1580901135676-0/employeeWithStatsFile/0_0_0.parquet]], selectionRoot=/home/runner/work/drill/drill/exec/java-exec/target/org.apache.drill.exec.sql.TestMetastoreCommands/dfsTestTmp/1580901135676-0/employeeWithStatsFile, numFiles=1, numRowGroups=1, usedMetadataFile=false, usedMetastore=true, filter=equal(`department_id`, 2) , columns=[`department_id`, `employee_id`]]]) : rowType = RecordType(ANY department_id, ANY employee_id): rowcount = 1155.0, cumulative cost = {1155.0 rows, 2310.0 cpu, 2310.0 io, 0.0 network, 0.0 memory}, id = 336734
 expected:<true> but was:<false> 

List of tests to fix:

• TestMetastoreCommands.testAnalyzeWithSampleStatistics;
• TestAnalyze.basic3.

Issue Links

• links to: GitHub Pull Request #1968