b628a1bfce
- Add hugepages and cpu_sets stanzas to HardwareProfile as the size and count of hugepages and the exact CPUs to pin for SRIOV are dependent on hardware. - Add sriov stanza to a node interface to specify vf_count and trustedmode. These will be passthrough values as Drydock doesn't configure SRIOV. - Add sriov information to the bootaction context so it can be written to disk on a deployed node if needed - Allow an interface configuration to be skipped when an interface has no defined network_link for things like SR-IOV interfaces. - Add kernel parameter reference support to access hardware profile information - Add unit tests - Update topology documentation for usage of HardwareProfile and kernel parameter references Change-Id: Iefd326f5c6fad19dbd21300ee249019a3dfd4848
drydock_provisioner
A python REST orchestrator to translate a YAML host topology to a provisioned set of hosts and provide a set of post-provisioning instructions.
See full documentation in docs/source/index.rst.
Required
- Python 3.5+
- A running instance of Postgres v9.5+
- A running instance of Openstack Keystone w/ the v3 API enabled
- A running instance of Canonical MaaS v2.2+
Recommended
- A running Kubernetes cluster with Helm initialized
- Familiarity with the AT&T Community Undercloud Platform (UCP) suite of services
Building
This service is intended to be built as a Docker container, not as a standalone Python package. That being said, instructions are included below for building as a package and as an image.
Virtualenv
To build and install Drydock locally in a virtualenv first generate configuration and policy file templates to be customized
$ tox -e genconfig
$ tox -e genpolicy
$ virtualenv -p python3.5 /var/tmp/drydock
$ . /var/tmp/drydock/bin/activate
$ pip install -r requirements-lock.txt
$ pip install .
$ cp -r etc/drydock /etc/drydock
Docker image
$ docker build . -t drydock
Running
The preferred deployment pattern of Drydock is via a Helm chart to deploy Drydock into a Kubernetes cluster. Additionally use of the rest of the UCP services provides additional functionality for deploying (Armada) and using (Promenade, Deckhand) Drydock.
You can see an example of a full UCP deployment in the UCP Integration repository.
Stand up Kubernetes
Use the UCP Promenade tool for starting a self-hosted Kubernetes cluster with Kubernetes Helm deployed.
Deploy Drydock Dependencies
There are Helm charts for deploying all the dependencies of Dryodck. Use them for preparing your Kuberentes cluster to host Drydock.
Deploy Drydock
Ideally you will use the UCP Armada
tool for deploying the Drydock chart with proper overrides, but if not you can
use the helm CLI tool. The below are overrides needed during deployment
- values.labels.node_selector_key: This is the kubernetes label assigned to the node you expect to host Drydock
- values.conf.dryodck.maasdriver: This is URL Drydock will use to access the MAAS API (including the URL path)
- values.images.drydock: The Drydock docker image to use
- values.images.drydock_db_sync: The Drydock docker image to use