[openstack-dev] [tripleo] Updates from the containers squad: Where are we at?
Fox, Kevin M
Kevin.Fox at pnnl.gov
Tue Jan 31 17:17:53 UTC 2017
Great to hear the containerization process is going well. :)
Can we set aside some time during the PTG to talk about the process of detecting stale/building fresh containers at the PTG? Its a problem that all the kolla deliverables would also like solved. So maybe we can work together on it?
Thanks,
Kevin
________________________________________
From: Flavio Percoco [flavio at redhat.com]
Sent: Tuesday, January 31, 2017 8:13 AM
To: openstack-dev at lists.openstack.org
Subject: [openstack-dev] [tripleo] Updates from the containers squad: Where are we at?
Greetings,
In the last month, there’s been no update on what’s been happening in the
containers effort for TripleO so I thought I’d drop an email to… well… provide
updates.
The work in the DFG has been broken down into more granular steps that are
easier to account for. However, I’d like to focus a bit more on the architecture
side of things first.
There have been some major changes in how containers are deployed within TripleO
and how services are containerized. I’ll try to describe those below but bear in
mind that some details may change a little bit and that this is still a work in
progress.
* The new architecture aims to support a hybrid deployment where it’s possible
to run some services on baremetal and others on containers. This has been done
to ease the upgrade process from baremetal nodes to a containerize node and it
doesn’t change the goal of having all services containerized. More details
about this new architecture can be found here[0]. Please, check out the README
file.
* There’s no heat agent container anymore. Instead, we’re installing puppet in
the base image for every service and then run puppet from the service
container. A bit of more details about this change can be found in this blog
post[0] (Thanks Dan)
* The old copy-json file that used to create the config.json for the kolla image
was replaced with a heat-hook. More about this on this patch.[2]
* Several patches for containerizing services have been proposed. TBH, we’ve not
done a great job at using a single topic BUT if you open the patch[0], you’ll
see a list of patches depending on it that propose the containerization of
several projects.
* There’s a non-voting CI job that currently only runs on patches that modify
files related to the containers work. You can read more about this on this
patch[3]. The goal is to make this job voting as soon as Pike starts.
* Perhaps more RDO specific but there’s also been great progress in the building
pipeline for containers. We’re using kolla images as our base container images
and the image build process is being integrated with the TripleO CI pipeline.
This means that new images will be created whenever there’s a promotion in
TripleO. We’d like to take this one step further for Pike and be able to build
images on every patch merged in the various projects, just like we’d do with
RPM packages. I wanted to mention this because I love the fact that this helps
us testing kolla images thoroughly and also give back to the community in some
way. More on this can be found here[4]
* Support for a containerized deployment has been added to tripleo-quickstart
and it’s being used for the CI job.
* The containers squad will start the work of documenting the current
architecture as soon as possible. As mentioned in the points above, there have
been several changes to the architecture that made starting the documentation
process not worthwhile. We can start documenting things now that we feel more
comfortable with the current version of the framework.
* We’re starting to work on a strategy for upgrades for baremetal nodes to
containerized nodes. Again, this work was waiting for the architecture to be
polished out a bit more and the composable upgrades work to mature. Now it
seems like a good time to kick this off.
Eventually (and likely soon) this will impact other squads and there will be
need of a joint work towards finalizing the containerization process. More
details will be provided as part of the documentation and in future emails.
There are still many things to do, though. There are services that still need to
be containerized, there are more CI jobs to be added and all the services that
have been containerized so far must also be tested in the overcloud. That being
said, I’m personally happy with the progress made so far and I hope you are as
well and that this email is useful to you.
Rock on,
Flavio
[0] https://review.openstack.org/#/c/416421/
[1] https://dprince.github.io/docker-puppet.html
[2] https://review.openstack.org/#/c/416420/
[3] https://review.openstack.org/#/c/423519/
[4] https://etherpad.openstack.org/p/rdo-kolla-build
--
@flaper87
Flavio Percoco
More information about the OpenStack-dev
mailing list