Jaeger with Grafana Tempo Backend

[robbydf]

Is there a guide or is it even feasible to use Grafana Tempo as a storage backend for Jaeger? We currently use Tempo but:

1. The UI (Grafana) is slow for traces that contain many spans making scrolling a jarring experience. This is my main concern. In contrast Jaeger UI seems fast.
2. There isn't a simple and short URL to share a trace id. In contrast, Jaeger has this.

My preference is to stay in Tempo but have the option to view the traces in Jaeger if needed.

[dosubot[bot]]

Hello @robbydf! I'm here to help you with your Jaeger and Grafana Tempo backend question. Let's work together to find a solution.

The Jaeger documentation does not specifically mention integration with Grafana Tempo as a storage backend. However, Jaeger supports custom or third-party storage backends through a gRPC-based Remote Storage API. This allows you to extend Jaeger's ecosystem with custom storage solutions that are not directly supported by the project ^[1][2][3][4].

To use a remote storage as Jaeger's storage backend, you would configure Jaeger to use grpc as the storage type and specify the remote gRPC server address. This setup might require additional development to create a compatible gRPC server that interfaces with Grafana Tempo.

If you are considering this approach, you might need to explore or develop a gRPC server that can translate between Jaeger's storage API and Grafana Tempo's API. This could potentially address your concerns about UI performance and trace sharing capabilities by allowing you to use Jaeger's UI with Tempo's data.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[yurishkuro]

I never tried it, and don't know how well it is supported, but Tempo includes an implementation of Jaeger Remote Storage API https://github.com/grafana/tempo/tree/main/cmd/tempo-query. It's a separate binary that you run as a gRPC server and point Jaeger query to it, and it will bridge Tempo DB as a compatible source for Jaeger. However, this binary does not implement the write path, only the query, so you would need to ingest data into Tempo directly. I think Tempo will accept OTEL protocol OTLP.

[robbydf]

Looks to be an undocumented component in Tempo for some reason... But thank you for pointing this out. We'll give it a try.

[yurishkuro]

If you make it work please update this ticket, we can include this on our docs.

[dannytest2709-droid]

“I’d like to ask, does Tempo have many bugs in actual use? Is it easier to use compared to Jaeger or SkyWalking?”

[dannytest2709-droid]

“I’d like to ask, does Tempo have many bugs in actual use? Is it easier to use compared to Jaeger or SkyWalking?”

[jkowall]

Tempo is not an officially supported backend for Jaeger, while you can use it this is not something the team here supports. There are no known bugs with Tempo as it's used extensively across the install base of Grafana Labs.

[jkowall]

1. The UI (Grafana) is slow for traces that contain many spans making scrolling a jarring experience. This is my main concern. In contrast Jaeger UI seems fast.

Tempo does not index the data, so free text search and recalling large spans will be slower on Tempo. Tempo does caching and other tricks to try to speed this up, but ultimately it's a different type of datastore and architecture.

This is not due to Grafana but the backend. Jaeger's supported backends use indexing or have very fast data recall making it more responsive.