Conversation
Bootstrap Kafka system tests
…nchronization for thread safety on println statements.
…hput logic to make it reusable.
…e intuitive behavior when targetThroughput is 0.
Verbose client
|
Probably worth clarifying that before this gets committed we'll want one final update to pin ducktape to a specific version, but there's no need push that ducktape release to PyPI until we're sure we have all the changes to ducktape we need to land this patch. We don't normally want people to have to checkout ducktape from its repository in order to run the system tests. |
|
As Ewen mentioned, I won't pin these changes to a ducktape release until just before merging in case updates to ducktape are requested while this pull request is being examined. To clarify what changes are present in this pull request: This initial pull request contains a small number of tests which starts but does not complete the process of porting/replacing existing system test code (https://cwiki.apache.org/confluence/display/KAFKA/Roadmap+-+port+existing+system+tests) It also has updates to Vagrant necessary for running the tests on a Vagrant cluster. Kafka codebase:
Vagrant:
tests/ directory:
|
There was a problem hiding this comment.
0.8.3-SNAPSHOT -> 0.8.3
Why? According to packaging standards (https://www.pypa.io/en/latest/peps/#pep426-metadata-for-python-software-packages-2-0), PEP440 should be used as a guide to versioning for pip and other packaging tools.
There was a problem hiding this comment.
Are we sure we want to change this now? We probably shouldn't even be publishing to PyPI except on releases when this would change anyway. In fact, shouldn't we just try to grab it out of build.gradle or wherever it normally is so we don't have version information duplicated in more locations than necessary?
There was a problem hiding this comment.
Agree with @ewencp - Apache is pretty strict about separating work-in-progress artifacts and public-consumable artifacts, and they have a bunch of legal reasons for that: http://www.apache.org/dev/release.html#releases
An alternative to grabbing version out of build.gradle is a small release script that will update the version everywhere at once.
|
FYI: git clone [email protected]:confluentinc/ducktape.git |
|
@granders In light of https://issues.apache.org/jira/browse/KAFKA-2316 this patch should probably be updated to use JDK7. Might require a couple of changes to the Vagrantfile in addition to updating the jdk6 references in the code added by this patch. |
…in dependencies on the Jackson JSON tools or argparse4j.
|
@junrao Would you mind looking through the changes introduced in the last commit (creating separate tools jar)? This avoids adding extra dependencies to the client jar as we discussed, but we will still need to fix the releaseTarGz and uploadArchives targets. |
|
kafka-trunk-git-pr #23 FAILURE |
|
kafka-trunk-git-pr #31 FAILURE |
|
kafka-trunk-git-pr #53 FAILURE |
|
kafka-trunk-git-pr #54 SUCCESS |
|
@guozhangwang, I notice that this was merged even though there is no LGTM in the PR. A number of reviewers were added to the commit message, so perhaps the LGTM was given somewhere else? Still, for clarity and transparency, it's important that people state in the PR that they are happy with it before it actually gets merged. A separate point is who is included as a reviewer in the commit message. I'll raise that in the mailing list. |
|
@ijuma you are right, it is my bad. I thought Jun / Gwen have gave it a +1 but I did not double check myself. |
Fix some metrics
…pache#70) Also moved KRaft mode above zk mode. Reviewers: Mickael Maison <[email protected]> Co-authored-by: Ismael Juma <[email protected]>
* fix(docker): add control-plane config to fix compose * chore(tools): add make targets for common gradle tasks
add mirror throttling
7 participants
Initial patch for KIP-25
Note that to install ducktape, do not use pip to install ducktape. Instead: