prodfiler-documentation

Frequently Asked Questions

I have questions not answered here! Where should I ask them?

On our community page.

What language runtimes does Prodfiler support?

The following programming languages are currently supported:

How much CPU and RAM will the agent consume?

The Prodfiler agent is designed to be used continuously in production and has a tiny performance footprint (< 1% CPU, ~200MB of RAM).

Can Prodfiler crash my kernel?

Prodfiler makes use of eBPF to load programs into the kernel. An eBPF program must pass the eBPF verifier, which ensures that the program is safe to run, before it’s loaded into the kernel. This means that even if the eBPF program is buggy, it can not crash the kernel.

How many machines can I deploy the agent to without your backend infrastructure melting?

There is a maximum number of projects (1) and a per-project host limit (20) for new accounts. Please contact us to have that limit raised.

Why are my stack traces not symbolised?

The Prodfiler agent does not require debug symbols in order to unwind stacks, but our backend does require them in order to symbolise stack traces in the UI. We automatically mirror, index and use the debug packages provided by several distributions (e.g. Ubuntu, Debian, Fedora and Amazon Linux), so if your traces are from applications in the official repositories then they should be automatically symbolised. If they are not, please let us know via the community page.

For custom binaries, or binaries that we have not indexed, you need to ship us your symbols. We provide a script to do so, which can be integrated with your CI/CD system if necessary. See here for details.

Filtering