Standalone Cinder Containerized using Docker Compose
Provides Block Storage as a service as part of the OpenStack Project. This project deploys Cinder in containers using docker-compose and also enabled the use of Cinder's noauth option which eliminates the need for keystone. One could also easily add keystone into the compose file along with an init script to set up endpoints.
The block-box uses OpenStack Loci to build a base Cinder image to use for each service. The examples use Debian as the base OS, but you can choose between Debian, CentOS and Ubuntu.
We're currently using Cinder's noauth option, but this pattern provides flexibility to add a Keystone service if desired.
Start by building the required images. This repo includes a Makefile to
enable building of openstack/loci images of Cinder. The
Makefile includes variables to select between platform (debian, ubuntu or
centos) and also allows which branch of each project to build the image from.
This includes master, stable/xyz as well as patch versions. Additional
variables are provided and can be passed to make using the -e
option to
control things like naming and image tags. See the Makefile for more info.
If you're going to utilize an external storage device (ie not using LVM), all you need to build is the base Cinder image. Set the variable in the Makefile to choose the Cinder Branch you'd like to use and Platform then simply run:
make base
You can also build an image to run LVM (NOTE: This is dependent on the base cinder image):
make lvm
To build both the base and lvm enabled image
make blockbox
All we're doing here is a docker build
utilizing the loci
Dockerfile and bindeps. Currently loci images are not published
regulary (although they will be in the future) so we require you
to build images before running docker-compose up.
For more information and options, check out the openstack/loci page on github.
NOTE The loci project is moving fairly quickly, and it may or may not continue to be a straight forward light weight method of building container Images. The build has been known to now work at times, and if it becomes bloated or burdensome it's easy to swap in another image builder (or write your own even).
Creates a base image with cinder installed via source. This base image is enough to run all of the services including api, scheduler and volume with the exception of cinder-volume with the LVM driver which needs some extra packages installed like LVM2 and iSCSI target driver.
Each Cinder service has an executable entrypoint at /usr/local/bin.
NOTE If you choose to build images from something other than the default Debian base, you'll need to modify the Makefile for this image as well.
This is a special image that is built from the base cinder image and adds the necessary packages for LVM and iSCSI.
You can of course build a cinderclient container with a cinder
entrypoint and
use that for access, but in order to take advantage of things like the
local-attach extension, you'll need to install the client tools on the host.
Before using, you must specify these env variables at least,
OS_AUTH_TYPE
, CINDER_ENDPOINT
, OS_PROJECT_ID
, OS_USERNAME
.
You can utilize our sample file cinder.rc
, then you can use client
to communicate with your containerized cinder deployment with noauth!!
Remember, to perform local-attach/local-detach of volumes you'll need to use
sudo. To preserve your env variables don't forget to use sudo -E cinder xxxxx
docker-compose up -d
Don't forget to modify the etc-cinder/cinder.conf
file as needed for your
specific driver. The current default setup should give you the ability to
quickly deploy a fully functional stand-alone cinder deployment with LVM.
If you'd like to add your own external driver, it's quite simple, and we've
included an example for adding an additional volume service to the base
deployment/compose. See the section below for more details.
Note: If you use cinder-lvm
image, you must guarantee the required
volume group which is specified in the cinder.conf
already exists in
the host environment before starting the service.
We don't do multi-backend in this type of environment; instead we just add another container running the backend we want. We can easily add to the base service we've create using additional compose files.
The file docker-compose-add-vol-service.yml
provides an example additional
compose file that will create another cinder-volume service configured to run
the SolidFire backend.
After launching the main compose file:
docker-compose up -d
Once the services are initialized and the database is synchronized, you can add another backend by running:
docker-compose -f ./docker-compose-add-vol-service.yml up -d
Note that things like network settings and ports are IMPORTANT here!!
You can use your own cinderclient and openrc, or use the provided cinderclient container. You'll need to make sure and specify to use the same network that was used by compose.
docker run -it -e OS_AUTH_TYPE=noauth \
-e CINDERCLIENT_BYPASS_URL=http://cinder-api:8776/v3 \
-e OS_PROJECT_ID=foo \
-e OS_VOLUME_API_VERSION=3.27 \
--network blockbox_default cinderclient list
docker run -d -p 5000:5000
-p 35357:35357
--link mariadb
--name keystone
--hostname keystone
-e OS_PASSWORD=password
-e DEMO_PASSWORD=password
-e DB_HOST=mariadb
-e DB_PASSWORD=password
keystone
docker run -d -p 5672:5672 --name rabbitmq --hostname rabbitmq rabbitmq
docker run -d -p 8776:8776
--link mariadb
--link rabbitmq
--name cinder-api
--hostname cinder-api
-v ~/block-box/etc-cinder:/etc/cinder
-v ~/block-box/init-scripts:/init-scripts
cinder_debian sh /init-scripts/cinder-api.sh
docker run -d --name cinder-scheduler
--hostname cinder-scheduler
--link mariadb
--link rabbitmq
-v ~/block-box/etc-cinder:/etc/cinder
cinder_debian cinder-scheduler
docker run -d --name cinder-volume
--hostname cinder-volume
--link mariadb
--link rabbitmq
-v ~/block-box/etc-cinder:/etc/cinder
cinder-debian cinder-volume