Remove open-ils.ingest service and related code

[Jeff Godin]

It appears that the open-ils.ingest service and related code are no longer used nor required.

The working branch referenced below removes the open-ils.ingest service and configs/tests related to it, as well as some import scripts which appear to have been obsoleted by the move to in-db ingest.

There may be related code which should also be removed, or there may be reason to keep some/all of what I have removed here.

This was an experiment. Discussion welcome!

[Jeff Godin (jgodin)]

Working branch is user/jeff/remove_ingest_cruft

In gitweb:
http://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/user/jeff/remove_ingest_cruft

[Galen Charlton (gmc)]

I haven't looked at the patch in detail yet, but generally, +1 to getting rid of open-ils.ingest.

[Jeff Godin (jgodin)]

If removed, we will probably need to determine the best means of ensuring that existing configuration files are updated to remove reference to open-ils.ingest. This could be as simple as release notes, or could even involve a shim library for an extremely graceful deprecation process (though that might only be delaying the pain).

[Galen Charlton (gmc)]

Something that occurred to me while thinking about this: one consequence of the way that Evergreen's Perl modules get installed is that at the moment active steps would have to be taken during upgrade to remove Ingest.pm. I think we should consider adjusting Open-ILS/src/perlmods/Makefile.in to do

Build install --uninst 1

Removing dormant sections is about the easiest thing I could envision a script that updates opensrf.xml automatically doing; starting such a script is worth considering IMO to reduce the amount of manual comparison that is required at the moment during upgrade.

If that's not favored, I think putting a note in the release notes is sufficient. It would be a nice enhancement for OpenSRF for it to have a way to check if all defined services have an underlying implementation, however.

I'm not in favor of leaving a shim library around; I think (without testing) that if one neglects to remove the service definition that the OpenSRF startup scripts will charge along blithely -- which is not a huge deal since nothing else depends on the service.

I'm assuming, by the way, that there's no compelling reason to backport the removal of open-ils.ingest.

[Jeff Godin (jgodin)]

I've rebased this branch against current master and done some testing. I've also added release notes recommending the removal of open-ils.ingest and the removal of OpenILS/Application/Ingest.pm from the Perl @INC path.

I researched and tested the suggestion of "Build install --uninst 1", and unfortunately it doesn't do what we want.

In testing, removal of Ingest.pm without removing open-ils.ingest from the config was not fatal -- an obvious error was logged, but startup of the remaining services was not affected. Removal of the config without removing Ingest.pm on disk should likewise be no issue.

Adding pullrequest for rebased branch user/jeff/remove_ingest_cruft_rebase located at:
http://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/user/jeff/remove_ingest_cruft_rebase

[Bill Erickson (berick)]

Passes the eyeball test. Passes a few running-server tests. 'make check' is happy. I'm not seeing any dangling references. Looks good to me. Merged to master with a tiny typo fix as icing. (Agreed with Galen there's no compelling reason to back-port). Thanks, Jeff!