otelcol.receiver timing out if downstream component offline

[smaddock]

What's wrong?

Currently, if an otelcol.exporter has errors, that causes the otelcol.receiver to time out. Receivers in Alloy should function regardless of whether the components they connect to are successful. The effect is then that our app fails because the in-app exporter expects a resilient receiver to be present. Instead, if an Alloy exporter has errors, those should be logged by Alloy, and not pass the instability up the chain.

Steps to reproduce

Build a simple otelcol pipeline in Alloy with a dud final endpoint and try to write to the receiver.

System information

Linux 5.10.0-33-amd64

Software version

alloy, version v1.5.1 (branch: HEAD, revision: dc8a365)

Configuration

otelcol.receiver.otlp "test" {
	http {}

	output {
		logs = [otelcol.exporter.loki.test.input]
	}
}

otelcol.exporter.loki "test" {
	forward_to = [loki.write.test.receiver]
}

loki.write "test" {
	endpoint {
		url = "https://nowhere/loki/api/v1/push"
	}
}

Logs

Jan 15 16:10:09 alloy[]: level=warn msg="error sending batch, will retry" component_path=/ component_id=loki.write.test component=client host=nowhere status=-1 tenant="" error="Post \"https://nowhere/loki/api/v1/push\": dial tcp: lookup nowhere on 1.1.1.1:53: no such host"
Jan 15 16:10:09 my-app[]: ERROR: Export failure: cURL error 28: Operation timed out after 10000 milliseconds with 0 bytes received for http://localhost:4318/v1/logs