Observe end-to-end latency

[binarylogic]

As part of finishing https://github.com/timberio/vector-product/issues/14, we need to track how long each event spends within each component, ultimately tracking end-to-end latency.

Before I outline requirements, I'd like to understand if this is possible at the topology level, or do we need to instrument this within each component?

cc @lukesteensen

[Hoverbear]

I'm going to coordinate with @bruceg on his e2e acking work to make sure our solutions compliment each other.

[jszwedko]

User requested this feature here: #9311 (reply in thread)

[NasAmin]

We need this too.
We are trying to track end-to-end latency for vector i.e.
application emits log --> vector agent collects it --> process and transform the log --> push to DD.

We are trying to measure the time it takes from first reading the application log from the k8s source and the DD sink. Looking at the internal_metrics, there seems to be no way of stitching this info even from multiple components 🙏

[NasAmin]

@jszwedko sorry for tagging you but is there any news on this? Is this on the roadmap at all?

[jszwedko]

@jszwedko sorry for tagging you but is there any news on this? Is this on the roadmap at all?

Hi @NasAmin !

This is something we are considering doing soon, but hasn't officially been scheduled yet.

[jszwedko]

Closing in favor of #14379