We're considering a cloud subscription, but why are metrics so much more expensive than traces and logs? We'd have to disable most of our exporters for it to be affordable.

I'd assume it's compute and space. The metrics come in at a high sample rate THEN the database also needs to calculate interval aggregates for it. I found the default OTel (for k8s / linux) pushed WAY more metrics than I cared for: https://signoz-community.slack.com/archives/C06T3T7QESD/p1752599329469339?thread_ts=1752573117.710589&cid=C06T3T7QESD

Ooh, thanks for the link! Will read through that before it expires.

The HTTP metrics was surprisingly heavy, so I ended up disabling it since I could write queries on trace spans. After that collectionInterval was easy enough to set to 1-2m for a 2-4x reduction.

Does the new External APIs feature work without them though?

I'm not sure about that. Better for a Signoz team member. One of them gave me an env-var to change the metric sample being a delta vs absolute.

k

I have trace sampling on, so I need http.server metric. In my case, enabling all tracing was costing us too much