Hi all,
please see inline.
Il 24/11/2016 15:19, Francesco Chicchiriccò ha scritto:
Hi Andrea,
first of all, let me thank you for such initiative of yours: the
container (and Docker) hype is higher than ever, and it'd be a pity
not to have Syncope joining the party.
AFAICT, I would see two distinct categories of Docker images:
1. a single image with core, console and enduser - which will require
an external DBMS for internal storage (another existing image?)
2. separate images for core, console and enduser - where the first
will require an external DBMS for internal storage (another image?)
and a JMS queue for OpenJPA clustering (another existing image?)
The idea is that the single image can be used for simple deployments,
where the others are meant to be orchestrated into a more powerful,
scalable architecture.
At the moment the best thing to do is to build the image from Dockerfile
and run a simple (sample) environment with docker-compose (actual
persistence is with PostgreSQL). Image is not so usable itself.
Documentation will be very important in this case.
Scalability feature must be developed, yet. But it would be a "killing"
feature!
We could develop some docker-compose.yml files and/or Dockerfiles
(images) that fit with teh different purposes that you explained.
If we will create and release (alongside with other artifacts) such
Docker files, then we might also need to discuss if (and how) to
publish the related images.
Publishing to Docker Hub seems to be the most effective way, but at
the moment it is very unclear if there is a standard ASF approach
about it: some images are available under the 'apache' organization
[3] (automated builds), other under specific organizations, but still
appearing managed by their respective PMCs [4] (for public releases
rather than automated builds).
Moreover, a recent discussion about this topic on legal is available:
[5].
My impression is that it would be safer to "just" release the Docker
files + instructions for building and orchestrating the resulting images.
About this I agree, the overhead of creating an image on specific
machine instead of downloading from registry could be feasible at the
moment. Also because it is difficult to create a generic image that fits
with all purposes.
Best regards,
Andrea
Regards.
On 23/11/2016 17:15, andrea wrote:
Hi all,
In the last months I've worked a bit on integration between Syncope
and Docker, in particular a way to distribute Syncope as Docker image.
In the repo located at [1] you can find my current work. The wiki is
at [2].
There are 1_2_X and 2_0_X branches respectively hosting Dockerfile to
build 1.2.8 and 2.0.1 images. I provided also the docker-compose.yml
files to build working examples. Obviously this files are only a
starting point for the n-th installation, that has to be tuned
according to specific requirements (properties, datasources,
hostnames, etc.)
I'm still doing some refactoring and refinements on both branches and
master branch, that is outdated. And obviously I'm updating the wiki.
Time to integrate this work into Syncope project?
Or at least we can create a registry containing this sample images?
Do we have proposals to improve and consolidate this work or to
improve usability of these images?
WDYT?
Best regards,
Andrea
[1] https://github.com/andrea-patricelli/syncope-docker/tree/2_0_X
[2]
https://github.com/andrea-patricelli/syncope-docker/wiki/How-to-start-Syncope-Docker-application
[3] https://hub.docker.com/r/apache/
[4] https://hub.docker.com/r/apachesling/sling/
[5] https://issues.apache.org/jira/browse/LEGAL-270
--
Andrea Patricelli
Tirasa - Open Source Excellence
http://www.tirasa.net/
PMC member at The Apache Software Foundation
Syncope
