[
https://issues.apache.org/jira/browse/CASSANDRA-14157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17209835#comment-17209835
]
Adam Holmberg commented on CASSANDRA-14157:
-------------------------------------------
Running on a resource-constrained VM reproduces almost 100% for me.
I believe this is just a problem with the test. It looks like the read repairs
are valid, occasionally being triggered by queries "under the hood" in cqlsh.
The test rightly expects the [insert and subsequent
select|https://github.com/apache/cassandra-dtest/blob/b117565b8f0096a3ed2af05fdec6e014a05788a1/cql_tracing_test.py#L80-L98]
at CL.ALL to produce no read repairs. However, running with trace in cqlsh
[causes additional queries getting the trace
data|https://github.com/apache/cassandra/blob/83e1e9e45193322f18f57aa7cc4ad31d9d5a152d/bin/cqlsh.py#L1044].
They will be run at the session level of CL.ALL, but the data is written at
CL.ONE. Therefore these queries can cause RR.
Rather than try to work around this in-place, my suggested change relocates the
test out of cql_tracing_tests and uses the driver directly (cqlsh is not the
thing we're testing here).
[patch|https://github.com/apache/cassandra-dtest/compare/master...aholmberg:CASSANDRA-14157?expand=1]
[ci|https://app.circleci.com/pipelines/github/aholmberg/cassandra?branch=CASSANDRA-14157]
(note that another read_repair_test is failing on trunk for a different reason
CASSANDRA-16148)
> [DTEST] [TRUNK] test_tracing_does_not_interfere_with_digest_calculation -
> cql_tracing_test.TestCqlTracing failed once : AssertionError: assert 0 == 1
> -----------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-14157
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14157
> Project: Cassandra
> Issue Type: Bug
> Components: Test/dtest/python
> Reporter: Michael Kjellman
> Assignee: Adam Holmberg
> Priority: Normal
> Labels: dtest
> Fix For: 4.0-beta3
>
>
> test_tracing_does_not_interfere_with_digest_calculation -
> cql_tracing_test.TestCqlTracing failed it's assertion once today in a
> circleci run. the dtests were running against trunk.
> Although it has failed once so far, a quick read of the comments in the test
> seems to indicate that the assertion failing this way might mean that
> CASSANDRA-13964 didn't fully fix the issue.
> {code:python}
> if jmx.has_mbean(rr_count):
> # expect 0 digest mismatches
> > assert 0 == jmx.read_attribute(rr_count, 'Count')
> E AssertionError: assert 0 == 1
> E + where 1 = <bound method JolokiaAgent.read_attribute of
> <tools.jmxutils.JolokiaAgent object at
> 0x7f62d4156898>>('org.apache.cassandra.metrics:type=ReadRepair,name=RepairedBlocking',
> 'Count')
> E + where <bound method JolokiaAgent.read_attribute of
> <tools.jmxutils.JolokiaAgent object at 0x7f62d4156898>> =
> <tools.jmxutils.JolokiaAgent object at 0x7f62d4156898>.read_attribute
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
