This is the working draft of the system design.
profefe consists of the following components
A service that receives profiling data in the format of pprof (
pb.gz). Collector persists the data in a plugable storage.
It also provides an API for querying and retrieving stored profiles.
The only storage implemented currently uses Badger.
Note: it’s tempting to split collector into two separate applications (collector and querier). That might inctease service’s scalability. But is the subject of future research.
An optional library, that can be integrated into an application instead of
net/http/prof. The goal of agent is to periodically scrap pprof data from the running application and to send the data to collector.
Below is the schema of profefe’s components interaction:
If integrating the agent into the application is impossible or undesirable, but the application already exposes
net/http/pprof server, the pprof data can be scrapped using an external mechanism, that will send the data to collector using its HTTP API.
In the following example, the agent was replaced with an external script that is running inside a cronjob. It periodically requests profiling data from application’s pprof server and sends it to collector: