[
https://issues.apache.org/jira/browse/FLINK-19654?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17214510#comment-17214510
]
Robert Metzger commented on FLINK-19654:
----------------------------------------
Looking at the PyFlink table test: It seems that the test has 2 parts
1st: start a cluster and run a number of jobs
2nd: start a Kerberos secured? flink on yarn on docker cluster to run 2 jobs.
In one example, the 1st part took 10 minutes, the 2nd part 14 minutes.
I propose to remove the 2nd part from the pyflink table test: Running Flink on
a Kerberos secured cluster is already tested separately.
Looking at the PyFlink DataStream test: It seems to do a lot of the same python
setup procedures (which take roughly 2 minutes).
Can't we do the DataStream test in the same test as the PyFlink Table test?
(because we have already python set up, and a cluster running?)
> Improve the execution time of PyFlink end-to-end tests
> ------------------------------------------------------
>
> Key: FLINK-19654
> URL: https://issues.apache.org/jira/browse/FLINK-19654
> Project: Flink
> Issue Type: Bug
> Components: API / Python, Tests
> Affects Versions: 1.12.0
> Reporter: Dian Fu
> Priority: Major
> Attachments: image (7).png
>
>
> Thanks for the sharing from [~rmetzger], currently the test duration for
> PyFlink end-to-end test is as following:
> ||test case||average execution-time||maximum execution-time||
> |PyFlink Table end-to-end test|1340s|1877s|
> |PyFlink DataStream end-to-end test|387s|575s|
> |Kubernetes PyFlink application test|606s|694s|
> We need to investigate how to improve them to reduce the execution time.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
