On 29/10/16 16:01, Arnaud Héritier wrote:
Let's forget the maven job. Our statistics about its usage are useless
as it was installed by default and we may hope that its usage is
decreasing...

AFAIK, basic usage of job types has always been available in the statistics. From the September CSV:

Job type                     Count     %
IvyModuleSet                    6311   0.1
WorkflowMultiBranchProject     17229   0.2
InheritanceProject             20407   0.2
WorkflowJob                   139976   1.5
MatrixProject                 306274   3.2
MavenModuleSet               1236525  13.1
FreeStyleProject             7725160  81.7


It remains that upgrading the java version of master is one thing.
But you also need to upgrade the java version used by agents which
aren't supporting an automated deployment (hello JNLP slaves on windows
...).

My problem is (and was for years when I administrated many
instances) the readibilty of the changelog/roadmap to understand when an
important change was done. I have no problem to release a java 8 version
tomorrow in a Jenkins 3.0 release. But in a 2.38 or 2.51.1 (fictifs
versions) will be always really strange for me. A minor release
shouldn't enforce you to change your infrastructure deployment (and
possibly break many jobs but it's a different story)

Jenkins has never used some sort of semantic versioning, so I don't think it would make a difference. In any case, things seemed to go fairly smoothly when Jenkins started to require Java 7, and the changes were clearly highlighted in the changelog, communicated with blog posts, and on Twitter.


Le samedi 29 octobre 2016, Baptiste Mathus <[email protected]
<mailto:[email protected]>> a écrit :


    2016-10-29 14:52 GMT+02:00 Arnaud Héritier <[email protected]
    <javascript:_e(%7B%7D,'cvml','[email protected]');>>:

        From my POV the problem is that we don't know if we are talking
        about 5% or 50% :(


    Well, since yesterday, we do know some things
    
<https://docs.google.com/spreadsheets/d/1_gsorbySktXNjMywDFZE0_eTm_r2zfPKdmq7AqWO2uQ/edit#gid=1770972416>.

      * Java 8 as runtime represents 52% of the 133505 instances out
        there (yes, it's ignoring the opted-out ones)
      * Maven-plugin is installed 118711 times (yes, ignoring on purpose
        the fact it was auto-installed up to 2.x)
      * So, *at most* it's 46% of the users base who *might *be impacted
        here, if they were to upgrade in the very short term (that is:
        while also still using an EoLed JDK7 durably).

    Now, looking at the trend of the number of installs of that plugin,
    my guts feeling is that is less than that. And if you add that to
    the trend of Java 7 JVM users (see link above), then it is most
    probably going down even quicker.

    An example I have in mind is my previous company, pretty sure they
    didn't upgrade to Jenkins 2.x yet, so they still contribute to that
    stats, but we migrated off that plugins years ago after so many
    headaches and hours spent to debug different behaviours between
    Maven CLI and using it in Jenkins.

    Correlating with
    
http://stats.jenkins.io/plugin-installation-trend/jenkins-version-per-plugin-version.json
    
<http://stats.jenkins.io/plugin-installation-trend/jenkins-version-per-plugin-version.json>
    I think we would have even more data, but seems like that data is
    wrong.
    Gonna find the one who contributed that and hit him (I know him
    pretty well).

--
You received this message because you are subscribed to the Google Groups "Jenkins 
Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jenkinsci-dev/9158b3dc-5e6c-c6cf-82db-560a168a49ff%40orr.me.uk.
For more options, visit https://groups.google.com/d/optout.

Reply via email to