6.5 KiB
Showing BST Microbursts in Grafana via Simulation
This document runs through the steps needed to launch OpenStack Monasca and broadview-collector, simulate microburst activity incoming from a broadview agent, and configure grafana 2.6 to display this data as a graph.
What we are aiming for is represented in the following image:
Grafana displaying BST microburst data
The steps provided in this document do not require anything other than what is provided in the broadview and monasca github repositories, and does not require a BroadView-capable switch. The exercise can be completed as a single node OpenStack cluster with minimal configuration via devstack.
Bringing up OpenStack Monasca and BroadView Collector
Follow the steps outlined in https://github.com/openstack/broadview-collector/blob/master/devstack/README.txt to bring up a minimal OpenStack cluster that includes Monasca. Grafana is a web-based monitoring UI that is integrated with Monasca.
Be sure to test out the setup by running:
python bst_report.py
You can veify the operation of the broadview-collector and Monasca setup by using the monasca command line client to verify that metrics have been recorded by the monasca service.
Directions for using the monasca command line client are found in the above README.txt file.
If you do not see data, check the settings in /etc/broadviewcollector.conf. Ensure that monasca is configured as described in the README.txt file.
Using grafana
To work with grafana, one needs to do two things:
- configure a datasource
- create a dashboard
Launching grafana
First, visit the openstack dashboard, and login. If you used the instructions that are in the README.txt file to create your local.conf file, the username will be "admin" and the password will be "secretadmin".
In the dashboard, open the panel named "Monitoring". Click on "Monitoring", and then "Overview". Then click on the button labeled "Grafana Home". A browser tab should appear which will contain the grafana UI, and you should be logged in.
Creating the Monasca Data Source
On the left side of the screen you should see a "Dashboards" and "Data Sources". Click on Data Sources. Then, at the top of the screen, click on "Add new".
Fill out the settings for the data sources as shown in the figure Grafana data source
Instead of supplying an auth token, you might want to simply click on the "Keystone Auth" checkbox. However, a bug at the time of writing this document existed with this setting. To obtain an auth token, run the following coomand:
curl -s -X POST http://localhost:35357/v2.0/tokens -d '{"auth": {"passwordCredentials": {"username":"mini-mon", "password":"password"}, "tenantName":"mini-mon"}}' -H "Content-type: application/json"
The above will generate output that contains a token, a portion of which is shown below:
{"access": {"token": {"issued_at": "2016-04-11T23:24:32.658304Z", "expires": "2016-04-12T00:24:32Z", "id": "7e5ac6a6c2d94b8aac870a32af125059", "tenant": {"description": null, "enabled": true, "id": "fb12e3d3fc534ebaa1aba7c73b573fce", "name": "mini-mon"}, "audit_ids": ["tq_3-j68SPOGZ06zVfhuww"]}, "serviceCatalog": [{"endpoints": [{"adminURL": "http://10.14.244.207:8774/v2.1/fb12e3d3fc534ebaa1aba7c7
Here, the token is the "id" field (7e5ac6a6c2d94b8aac870a32af125059). Copy and paste this value into the Token text field.
The Url field corresponds to the IP address and port that the Monasca API is listening on. Make sure to change the IP address, and if necessary, the port.
Use the "Test Connection" button to verify the data source is properly configured, and then click on Save.
Starting the Simulator
At this point, it will be good to start the simulator to generate simulated microburst activity. Go to where broadview-collector has been cloned from github by devstack (likely /opt/broadview-collector), or clone it yourself. Then, cd into broadview-colle4cotr/broadview_collector/tools. Edit the script burst.py to set the host and port variables to the IP address and port that the collector is running on, then run the following in a bash window:
$ while true; do sleep 90; python bst_burst.py; done
This will get data flowing into the collector and then into the monasca databases.
Creating a Dashboard
A dashboard is where you display graphics or tables of metrics that are available in a selected datasource. The Monasca plugin, in association with grafana, will automatically determine the set of metrics available in the datasource for display, and provide UI that allows you to drill down on metrics based on name, dimensions, and other search criteria. The best way to experiment is perhaps to create a graph dashboard, then add panels, using the UI to add queries to the panel for each data you wish to view.
To get started, click on Dashboards, then click on the button labeled "Home" in the upper left corner. A dialog will display, with three buttons at the botton. Click on the one labeled "+New". A green bar will display, click on it and a pullright menu will display. From this menu, select "Add Panel->Graph". A graph will display, and below it will be an editor which you can use to add queries. Each of the queries will be displayed in the graph.
In the righthand side of the editor will be a button labeled "Grafana" This button is used to select a datasource. Click on the button and select "broadview bst", which is the datasource we created above. See the screenshot below.
Using the editor, set Function to "none", select "broadview.bst.device" as the Metric, and set the Group By Time text field to 5000.
Then, click on the "+Query" button, and add a second stat by doing the following:
- set Function to "none",
- Metric to "broadview.bst.egress-cpu-queue", and
- Group By Time to 5000.
In the Dimensions section, configure the following query "stat=cpu-buffer-count". The screen should look something like the following (depending on what data has been transmitted to the collector):
To make the graphic update frequently, use the controls that are in the upper right hand corner of the grafana dashboard viewer. They can be used to set the viewing window (e.g., last 6 hours, last 30 minutes) and the refresh frequency (10 seconds, 1 minute, etc.)