[jira] [Commented] (ACCUMULO-2229) Make init.d scripts get into the assembly in a more maveny way

["Christopher Tubbs (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ACCUMULO-2229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13880527#comment-13880527 ] 

Christopher Tubbs commented on ACCUMULO-2229:
---------------------------------------------

I think it's useful to put them with the service they manage, but I think it's overkill to create additional artifacts, and very much not necessary. They should probably be put in src/scripts/init.d or src/scripts/systemd or whatever, per module, to keep them with the service they manage. We can include it in the RPM/DEB packages right there... no need to add dependencies from siblings.

These init scripts are possibly also tightly coupled with the package (RPM for RHEL6/CentOS6, RPM for Fedora/RHEL7/CentOS7, Ubuntu) in which they are included, and it would get quite crazy if we created separate artifacts for each variant. A generic SysVinit script is available today... but that may not be the case in the future, as we may want to package upstart and systemd scripts, and I'd hate to start a precedent of making each of these separate artifacts... especially since they are likely to be 1-to-1 with the RPM/DEBs we build.

I don't see a problem continuing to access the (currently very generic) SysVinit scripts from the sibling module for the generic binary tarball. If we end up making systemd-based RPMs in the future (RHEL7/CentOS7, for instance), we might want to change the per-module scripts to systemd, but keep SysVinit scripts in the assembly module itself for generic systems.

In short, what I'm saying is... keep the scripts local to the RPM/DEB packaging, and reference them from the sibling module as long as they are generic enough to work in the generic binary tarball... and if they aren't generic enough for the tarball, put them in the assembly module.

> Make init.d scripts get into the assembly in a more maveny way
> --------------------------------------------------------------
>
>                 Key: ACCUMULO-2229
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-2229
>             Project: Accumulo
>          Issue Type: Bug
>          Components: build
>    Affects Versions: 1.6.0
>            Reporter: Michael Berman
>            Priority: Minor
>
> (forked from ACCUMULO-1983)
> For 1.6 the init.d scripts were moved into the module for the corresponding service rather than all being piled into the assemble module.  To get them into the assembly, the scripts are just copied by path out of assemble's siblings.  This is simple and it's easy to see what's going on when looking at the pom, but it definitely violates maven best practices (don't reference "..").  I think if we want to keep the init.d scripts with their corresponding modules, the maveny way to do it would be to declare the init.d script as an artifact of each module (of type "init.d" or something), and then declare them as dependencies of the packager, which could then use the copy-dependencies goal to get them into the assembly. It's more lines of pom and possibly more opaque as far as figuring out where each file is coming from, but it would be more portable and less sensitive to module rearrangements in the future.
> Is this a good idea?  Is it pom overkill?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)