kafka-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ewen Cheslack-Postava (JIRA)" <j...@apache.org>
Subject [jira] [Resolved] (KAFKA-1771) replicate_testsuite data verification broken if num_partitions > replica_factor
Date Fri, 30 Oct 2015 22:29:27 GMT

     [ https://issues.apache.org/jira/browse/KAFKA-1771?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Ewen Cheslack-Postava resolved KAFKA-1771.
    Resolution: Invalid

No longer valid since the old system tests have been removed.

> replicate_testsuite data verification broken if num_partitions > replica_factor
> -------------------------------------------------------------------------------
>                 Key: KAFKA-1771
>                 URL: https://issues.apache.org/jira/browse/KAFKA-1771
>             Project: Kafka
>          Issue Type: Bug
>          Components: system tests
>    Affects Versions:
>            Reporter: Ewen Cheslack-Postava
>            Assignee: Ewen Cheslack-Postava
>         Attachments: kafka-1771.wip.patch
> As discussed in KAFKA-1763,   testcase_0131,  testcase_0132, and testcase_0133 currently
fail with an exception:
> {quote}
> Traceback (most recent call last):
> File
> "/mnt/u001/kafka_replication_system_test/system_test/replication_testsuite/
> replica_basic_test.py", line 434, in runTest
> kafka_system_test_utils.validate_simple_consumer_data_matched_across_replic
> as(self.systemTestEnv, self.testcaseEnv)
> File
> "/mnt/u001/kafka_replication_system_test/system_test/utils/kafka_system_tes
> t_utils.py", line 2223, in
> validate_simple_consumer_data_matched_across_replicas
> replicaIdxMsgIdList[replicaIdx - 1][topicPartition] = consumerMsgIdList
> IndexError: list index out of range
> {quote}
> The root cause seems to be kafka_system_test_utils.start_simple_consumer. The current
logic seems incorrect. It should be generating one consumer per partition per replica so it
can verify the data from all sources, but it currently has a loop involving the list of brokers,
where that loop variable isn't even used.
> But probably a bigger issue is that it's generating multiple processes in the background.
It records pids to the single well-known entity pid path, which means only the last pid is
saved and we could easily leave zombie processes if one of them hangs for some reason.

This message was sent by Atlassian JIRA

View raw message