[
https://issues.apache.org/jira/browse/SOLR-11245?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17180794#comment-17180794
]
David Smiley commented on SOLR-11245:
-------------------------------------
I'm very inclined to think Solr should maintain its own Dockerfile. A rough
look at the PR looked fine but what's not clear are each delta from
"docker-solr" which is where you obtained this, of course. It's likely out of
date today. I propose re-doing where we start with one commit which is merely
a direct copy of the docker-solr version, and then a subsequent commit where
you make changes to accommodate this PR "living" here vs being external.
*But* wait until we (as a project) are ready; I'm not sure we're ready yet....
A quick 'n dirty copy is easy short-term but creates a large immediate
tech-debt (fork of scripts), and Solr doesn't need more tech-debt!
I think the real work is figuring out how to maintain both or how to live with
just one, or some sort of hybrid. Decisions, then work to migrate. A separate
docker-solr and the scripted/templeted way it works allows easy changes that
span all releases (e.g. the jattach addition). We might simply forgo that
advantage – latest & greatest always requires the latest release only. I can
support that in the name of maintenance / simplicity. Then, figure out how to
migrate the Docker level tests over to here. Maybe we assume gradle (9x)?
Figure out how CI will run them.
I'm not sure if there are any ASF matters / limitations pertaining to Docker,
and likewise there are any Docker.com limitations on how docker-library is run.
> Cloud native Dockerfile
> -----------------------
>
> Key: SOLR-11245
> URL: https://issues.apache.org/jira/browse/SOLR-11245
> Project: Solr
> Issue Type: New Feature
> Components: Build
> Affects Versions: 6.6
> Reporter: jay vyas
> Assignee: Jan Høydahl
> Priority: Major
> Labels: docker
> Time Spent: 10m
> Remaining Estimate: 0h
>
> SOLR Should have its own Dockerfile, ideally one that is cloud native (i.e.
> doesn't expect anything special from the operating system in terms of user
> IDs, etc), for deployment, that we can curate and submit changes to as part
> of the official ASF process, rather then externally. The idea here is that
> testing SOLR regression, as a microservice, is something we should be doing
> as part of our continuous integration, rather then something done externally.
> We have a team here that would be more then happy to do the work to port
> whatever existing SOLR dockerfiles are out there into something that is ASF
> maintainable, and cloud native, and easily testable, as well.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
