Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/20514 )

Change subject: KUDU-3507 Fix mini ranger port detection
......................................................................


Patch Set 2:

(2 comments)

http://gerrit.cloudera.org:8080/#/c/20514/2//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/20514/2//COMMIT_MSG@10
PS2, Line 10: lsof might handle 0.0.0.0 as ipv6
> The message is very misleading. Originally I thought that the ip address is
Thank you for taking time to write down those details.  I didn't do all the 
steps in the recipe, but I just realized that it would be much simpler to 
instruct the JVM to use IPv4 instead when both IPv4 and IPv6 are available.  
BTW, similar approach is used for the HMS-related tests (see mini_hms.cc).

Overall, the following small patch fixed the problem in my test environment:
  https://gerrit.cloudera.org/#/c/20582/

Could you apply it and see if it helps to address the issue in your test 
environment as well?  If it helps, maybe you could accommodate that simpler 
approach in this patch?


> This happened, because this is not the API port.

Yes, indeed -- it seems that's the Tomcat's shutdown port.  I think we should 
just disable it explicitly, so the process doesn't open that one since the test 
isn't using it anyway.


http://gerrit.cloudera.org:8080/#/c/20514/2/src/kudu/util/test_util.cc
File src/kudu/util/test_util.cc:

http://gerrit.cloudera.org:8080/#/c/20514/2/src/kudu/util/test_util.cc@584
PS2, Line 584:     // lsof exits with 1 even if there is no data:
             :     // https://github.com/lsof-org/lsof/issues/128
             :     // (-Q option would solve it, but its only available on new 
systems
             :     // like ubuntu 23.04)
> Sorry, I am wrong on the "completely broken" part. If the <PID> is filtered
OK, even if lsof returns 1 instead of 0 with empty output per the specified 
flags, I guess it's not 100% relevant to this  issue of failing a particular 
test.  At least I haven't seen a single failure of the test because of this 
peculiarity in lsof's behavior at the machines where the test scenario fails 
every time it runs.

If you want to address the issue with that lsof behavior, could you please do 
so in a separate patch?  Keeping semantically/logically separate matters in 
separate patches pays off later if trying to track down reasons why this or 
that change has been done in the code, and it helps in cherry-picking relevant 
patches in maintenance branches, if needed.

Thanks!



--
To view, visit http://gerrit.cloudera.org:8080/20514
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I89c3409e9efd53a821a6d4244a35564e134323bb
Gerrit-Change-Number: 20514
Gerrit-PatchSet: 2
Gerrit-Owner: Zoltan Martonka <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Zoltan Martonka <[email protected]>
Gerrit-Comment-Date: Tue, 17 Oct 2023 00:05:15 +0000
Gerrit-HasComments: Yes

Reply via email to