Is Grafana Agent is necessary?

Is Grafana Agent is necessary for trace collection by tempo? According to “Gettting started” page it is not, it is marked as an optional.

But I cannot find any examples of such configuration, also other page says it is needed: Set up and use tracing | Grafana Cloud documentation

It is relevant for us, as in our use case: we would like to trace serverless cloud functions on GCP, and actually we would need to setup additional infrastructure just to setup agent instance with correct networking - so that they will be able to communicate with Agent instance.

Side question: I also tried to find some guidelines what are requirements for Agent host, but couldn’t find any.

Thank you in advance.
Best Regards

Hi @marcinpolak.

It’s optional, yes. More broadly, tracing pipelines are optional but recommended, specially with big deployments. The Grafana Agent can be run as a tracing pipeline, but there are other tools out there as well, such as the OTel collector.

I think that doc is misleading :confused:. As I said, it’s not a must, only recommended. The Grafana Agent is a simple and quick way to send traces to Grafana Cloud. It uses the supported protocols/formats and auth is easy to configure.

For tracing, the Grafana Agent doesn’t really need anything special. It only need to be able to listen to TCP/UDP/HTTP/gRPC (depending on the protocol and instrumentation used), and forward gRPC traffic (it can also work with HTTP if not using multi-tenancy or Grafana Cloud). In kubernetes, it is commonly run as a deployment. Were you thinking on something in particular?

Thank you @mariorodriguez

We would like to evaluate Grafana Cloud for tracing our process that is running on GCP cloud functions.
For now I’m looking for simplest way to start collecting traces to play with Tempo.
Therefore we would like to know is way directly send traces from app to Grafana Cloud.