[IMPALA-9678] Dockerised tests with USE_CDP_HIVE=true crash with orc-metadata-utils.cc DCHECK

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Duplicate
• Affects Version/s: Impala 4.0.0
• Fix Version/s: Impala 4.0.0
• Component/s: Backend
• Labels: broken-build
• Target Version: Impala 4.0.0
• Epic Color: ghx-label-6
• Assignee: Joe McDonnell
• Reporter: Joe McDonnell
• Created: 20/Apr/20 16:39
• Updated: 28/Apr/20 00:18
• Resolved: 28/Apr/20 00:18

Description

I ran a ubuntu-16.04-dockerised-tests-cdp-hive on master, and all of the Impalads crashed with this message during end-to-end tests:

F0419 03:32:48.973024 31936 orc-metadata-utils.cc:160] 444f457c178cfeff:de098d6100000001] Check failed: i == col_path.size() - 1 (0 vs. 1) 
F0419 03:32:48.974320 31935 orc-metadata-utils.cc:160] 444f457c178cfeff:de098d6100000001] Check failed: i == col_path.size() - 1 (0 vs. 1) 
F0419 03:32:48.974320 31935 orc-metadata-utils.cc:160] 444f457c178cfeff:de098d6100000001] Check failed: i == col_path.size() - 1 (0 vs. 1) F0419 03:32:48.981420 31937 orc-metadata-utils.cc:160] 444f457c178cfeff:de098d6100000001] Check failed: i == col_path.size() - 1 (0 vs. 1) 
F0419 03:32:48.974320 31935 orc-metadata-utils.cc:160] 444f457c178cfeff:de098d6100000001] Check failed: i == col_path.size() - 1 (0 vs. 1) F0419 03:32:48.981420 31937 orc-metadata-utils.cc:160] 444f457c178cfeff:de098d6100000001] Check failed: i == col_path.size() - 1 (0 vs. 1) F0419 03:32:48.983081 31938 orc-metadata-utils.cc:160] 444f457c178cfeff:de098d6100000001] Check failed: i == col_path.size() - 1 (0 vs. 1) 

It's coming from this DCHECK:

    if (table_col_type->type == TYPE_ARRAY) {
...
    } else if (table_col_type->type == TYPE_MAP) {
...
    } else if (table_col_type->type == TYPE_STRUCT) {
...
    } else {
      DCHECK(!table_col_type->IsComplexType());
      DCHECK_EQ(i, col_path.size() - 1);            <--------
      RETURN_IF_ERROR(ValidateType(*table_col_type, **node));
    }

https://github.com/apache/impala/blob/master/be/src/exec/orc-metadata-utils.cc#L160

https://jenkins.impala.io/job/ubuntu-16.04-dockerised-tests-cdp-hive/4/

This matches the GVO failures that we saw when trying to set USE_CDP_HIVE=true by default (https://gerrit.cloudera.org/#/c/15710/). It seems to reproduce consistently, but only on the dockerised job.

Issue Links

• is fixed by: IMPALA-9685 Full-ACID support breaks in LocalCatalog mode