Test dynamic column names in resultsets with less partitions created

[jeeminso]

Summary of the changes / Why this is an improvement

Improves #344.

Checklist

• Link to issue this PR refers to (if applicable): Fixes #???

[jeeminso]

Hi @matriv , I think my previous approach created more partitions/shards than it needed to and caused time outs / weird state of the cluster leading to such exceptions. Could you take a look at this?

Also do you think it is worth to investigate how the null rows got inserted?

  File "/var/lib/jenkins/workspace/CrateDB/qa/crate_qa/tests/bwc/test_upgrade.py", line 326, in assert_data_persistence
    for name in row[0].keys():
                ^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'keys'

[matriv]

Sorry I don't get your approach now, why do you need a new table, but keep a dynamic object on the partitioned table? Also why did you restrict the inserts to only 2 versions?

[jeeminso]

Sorry I don't get your approach now, why do you need a new table, but keep a dynamic object on the partitioned table? Also why did you restrict the inserts to only 2 versions?

Those are results of the reversion, it would be clearer to check the last commit of the two.

[matriv]

Thank you @jeeminso, I was confused before.

[matriv]

retest this please

[jeeminso]

The exception due to noneType is gone with the approved changes but still timing out, even removing most of the test added originally did not help. I am guessing it is some kind of env issue.

[matriv]

retest this please

[jeeminso]

retest this please

[jeeminso]

retest this please

[jeeminso]

retest this please