browbeat/ci-scripts
jkilpatr c1e59ddb19 Fix string literals
'' is different from "" in Bash. Leaving this up for
whatever the next problem I find is so that I can stop
polluting the commit history with tiny fixes.

Change-Id: Ia1d53670347a3862c743a4a74b0f0d3eda62dbd8
2016-12-10 13:55:01 +00:00
..
config Set oooq CI to use ssh-dns option 2016-12-07 16:13:46 +00:00
tripleo Fix string literals 2016-12-10 13:55:01 +00:00
README.rst Add oooq Ansible Role Support 2016-11-30 20:34:31 +00:00

Table of Contents

CI Structure

For an example Jenkins configuration see this job

If you would like to make your own CI job add your CI script to this directory and invoke it as minimally as possible on the Jenkins end, this will help us keep script changes in the repository and better test them before merging.

Script Documentation

Install and Check

Currently the main CI script that is run against every commit submitted to the Openstack Gerrit. For each test a fresh Openstack instance is deployed using TripleO Quickstart, Browbeat is then installed. Both of these happen regardless of what was included in the commit. Workload tests are run only if a file diff between the commit and Browbeat master contains the workload name. Success is defined as all processes in the script exiting with exit code 0, note Browbeat will return zero if a test fails its SLA or otherwise fails in a manner that's not total.

To add an additional workload to the script add the workload name to the tools loop near the bottom of the file.

for tool in rally perfkit shaker <tool name>; do

Then add configuration details that run all functions of the added task or plugin to the browbeat-ci.yaml file in ci-scripts/config.

You can view the output of this job here

Invoking Locally

To run tripleo/install-and-check.sh using your local machine as the driver for a TripleO Quickstart / Browbeat deployment create an empty directory to use as your workspace and point virthost at a machine running CentOS 7+ or RHEL 7+ with at least 32gb of ram.

$ export WORKSPACE=<your empty directory>
$ export VIRTHOST=<deployment machine hostname>

Navigate to the workspace directory

$ cd $WORKSPACE

Clone the required repositories

$ git clone https://github.com/openstack/browbeat
$ git clone https://github.com/openstack/tripleo-quickstart/
$ git clone https://github.com/redhat-openstack/ansible-role-tripleo-inventory

Install the Ansible roles from Github into the virtual environment, as well as a few Python packages

$ virtualenv --no-site-packages $WORKSPACE
$ source $WORKSPACE/bin/activate
$ cd $WORKSPACE/ansible-role-tripleo-inventory/
$ python setup.py install
$ cd $WORKSPACE/tripleo-quickstart
$ python setup.py install
$ pip install --upgrade ansible netaddr

Install the package dependencies, if you're nervous about using root just look inside of quickstart.sh, these are very generic and you might already have all of them installed.

$ sudo bash $WORKSPACE/tripleo-quickstart/quickstart.sh --install-deps

Finally invoke the script and settle in, this command will take about two hours to complete and will place all the relevant ssh credentials and other information to access your instance once the run is complete in the workspace directory.

$ bash $WORKSPACE/browbeat/ci-scripts/tripleo/install-and-check.sh mitaka delorean minimal periodic

Browbeat as a Quickstart Extra

TripleO Quickstart provides an extensible interface to allow "Extras" to add to to its core functionality of generating an entierly virtual Openstack Deployment using TripleO. The focus of this script is to deploy barmetal clouds in CI for effective and extensible automated benchmarking.

Invoking Locally

Please read The Extras Documentation for a general background on how TripleO Quickstart Extras operate. Please also reference The Baremetal Environments Documentation if you need to configure your job to run on baremetal.

Browbeat provides two playbooks for use with Quickstart quickstart-browbeat.yml and baremetal-virt-undercloud-tripleo-browbeat.yml the first playbook is for deploying an entierly virtual setup on a single baremetal machine. The second playbook creates a virtual undercloud on a undercloud host machine and deploys a baremetal overcloud as configured by the users hardware environment.

Dependencies for this script (at least for Fedora 25) are

sudo dnf install ansible git python-virtualenv gcc redhat-rpm-config openssl-devel

To run virtual TripleO Quickstart CI set the following environmental vars and run quickstart-virt.sh this will create a TripleO environment and run a short Browbeat test. Since this is a all virtual setup it is not suggested for serious benchmarking.

export WORKSPACE={TripleO Quickstart Workspace} export RELEASE={release} export VIRTHOST={undercloud-fqdn}

pushd $WORKSPACE/browbeat/ci-scripts/tripleo

bash quickstart-virt.sh

To run the baremetal CI follow the requisite steps to setup a hardware environment (this is nontrival) then create a workspace folder and clone TripleO Quickstart and Browbeat into that workspace. Set the variables below and then run microbrow.sh.

export WORKSPACE={TripleO Quickstart Workspace} export HW_ENV={hw-env} export RELEASE={release} export GRAPH_HOST={Graphite + grafana host} export GRAFANA_USER={username} export GRAFANA_PASS={password} export CLOUD_NAME={cloud-name} export BENCHMARK={benchmark config file ex browbeat-basic.yaml.j2} export ELASTIC_HOST={elastic host} export VIRTHOST={undercloud-fqdn}

pushd $WORKSPACE/browbeat/ci-scripts/tripleo

bash microbrow.sh

Configurable Options

By default a cloud will be setup and a very basic benchmark will be run and all results will be placed only in the browbeat/results folder on the virtual undercloud. If configured to use Elasticsearch metadata and benchmarks results will be inserted into Elasticsearch for easier visualization and storage. If Graphana is enabled performance metrics will be gathered from all cloud nodes and stored into the configured graphite instance to be processed by the Grafana dashboards created using the given username and password. These dashboards will be automatically overwritten each run to reflect the number of nodes in your cloud and other changes that may occur between runs.