Starting with Grafana 9.3 Parca will be a core plugin in Grafana, meaning it is available out of the box without having to install any plugins to utilize Parca. This is huge for lowering the bar of entry to the Parca project!
Aside from the core Parca plugin, a flamegraph panel was also added, that can be used by any datasource. It's great to see Grafana acknowledging the power of continuous profiling and the Parca project as a whole. To learn more about Parca and how to best make use of continuous profiling, check out our other blog posts.
Also congratulations to our friends at Grafana for the other amazing launches they announced at ObservabilityCon!
Trying out Parca in Grafana
Since the flame graph panel and Parca data source haven't landed in a Grafana release yet, you need to run Grafana with a feature flag called `flameGraph`.
Running the latest main branch in Docker with the feature enabled looks something like:
docker run --name grafana --rm -p 3000:3000 -e GF_FEATURE_TOGGLES_ENABLE=flameGraph grafana/grafana:main
It'll run the current main branch of Grafana and enable the flameGraph feature flag. Once the container is up and running you can access the local instance on port 3000: http://localhost:3000
Username and password are both admin.
Connecting to demo.parca.dev
Once logged into Grafana go into the Data source settings on the left hand navigation:
A screenshot of Grafana's menu to find data source configurations
You want to add a new data source:
A screenshot from within Grafana showing the add datasource button
Scrolling down you should eventually see the Parca data source:
Screenshot of the Parca core plugin in Grafana
Clicking on the row will add the data source and show its configuration next:
A screenshot of a Parca datasource configuration within Grafana
As you can see, the only thing that needs to be done is adding the URL: https://demo.parca.dev
Next click on "Save & test" and once you got the success confirmation you can open Explore:
Query Parca through Grafana Explore
On the Explore page you need several steps to get a similar result as shown in the screenshot:
- Select the Parca data source at the very top
- Select the profile type (parca_agent_cpu - samples)
- Write label matchers to slice and dice the processes you want to look at
- Finally, actually run the query by clicking on the blue button at the top right.
We recommend that starting with Grafana 9.3 or higher the core plugin is used, and for the time being we will maintain the separate plugin for versions lower than that. Once Grafana 9.3 has reached a critical mass, we will focus on the core plugin.