OpenTelemetry Bridge

[isbm]

Hey!

So if you've ever wanted to take a Perfetto trace and send it to an OpenTelemetry
collector (Datadog, Grafana, Honeycomb, etc.), then I just built a... thing that
does exactly that (because I need it 😉).

General Idea

1. Feed it a .pftrace file
2. Spawn trace_processor, maps every Perfetto table (sched slices,
   thread states, ftrace events, wakeups, counters, etc.) into OTel log/trace
   records with structured attributes
3. Store them in a durable append-only .logjet container (OTLP-native,
   replayable, checksummed, "indestructible" transfer wire serialisation protocol for OTel)
4. Forward to any OTel collector, or inspect offline (profit, basically)

Two Modes

Mode	How
SQLite (default)	trace_processor export sqlite to read via sqlite
RPC (no temp files)	trace_processor server stdio to protobuf over pipe

Yes, Demos

Quick demo on Linux (needs sudo for ftrace):

git clone https://github.com/google/perfetto
git clone https://github.com/tinythings/logjet
cd logjet
make dev && ./demo/perfetto/build-perfetto.sh
./demo/perfetto/perfetto-to-logjet/run-demo.sh

It will build Perfetto on sibling directory. Then it will records 5 seconds
of CPU scheduler events, imports them as OTel records, and opens an interactive
not fancy TUI viewer, but you can see that ~30 seconds end-to-end. It will look
something like this:

Get It

• Plugin: plugins/perfetto-ingest/ (cdylib, loaded by the logjet daemon at runtime)
• RPC client: manual protobuf encode/decode for the trace_processor stdio protocol
• Works on any Linux with ftrace; Perfetto is bundled and built locally

Not an official Perfetto project, just something I needed so I built it.
If you try it and it breaks, I'd love to hear about it (I also accept PRs BTW 😛).

More on Logjet: https://github.com/tinythings/logjet

Hope it helps!