Continuous Profiling is awesome! It helps in many situations, OOMKills (out-of-memory kills), improving latency, as well as saving cost for your infrastructure.
Therefore, we are super excited to introduce a Parca plugin for Grafana!
Why create a Grafana plugin?
When we set out to create Parca, originally called conprof, we wanted a continuous profiling tool that was well-integrated into the broader ecosystem, especially with Prometheus and Kubernetes. From the beginning the idea was to extend the known "Three Pillars of Observability", yet meeting users where they are, so naturally, integrating Parca into the Grafana dashboards would be an obvious choice in the long run.
In the past months Parca has matured a lot and the APIs have never been more stable, although still improving week after week, we felt like it’s the right time to tackle what so many have asked for, both on the internet and at conferences like KubeCon.
How to use Parca in Grafana today
Here are the Grafana plugin repository links for the plugins:
You can checkout Grafana's plugin installation guide for more details.
Once plugins are installed, it is pretty straightforward to use them. Like any other Grafana plugin, you need to create a Parca data source and then add panels that use Parca Flame graph to visualize the profiles from Parca data source.
Please refer to the docs for more details.
Try the hosted demo
Similar to demo.parca.dev/ running a version of Parca that everyone can always try, we have also setup a hosted version of Grafana with the Parca plugins. It’s pre-configured to use that same parca demo instance and you can take a look at it: demo.parca.dev/grafana
Please reach out to use to give us feedback on how to make this better.
You can always join the Parca Discord to ask questions.
Additionally, next week Matthias, Kemal and Frederic are going to be at KubeCon North America 2022 in Detroit. Feel free to reach out or directly approach us to talk about this Grafana Parca plugin or for any other Observability related topic!