Parseable cannot parse OTLP data without `schemaUrl`

[MatthiasRoels]

When wiring up Parseable (OSS) with OpenTelemetry (configured with a TargetAllocator to be able to scrape Prometheus ServiceMonitor resources), I get the following error:

parseable::handlers::http::modal::utils::ingest_utils:152: Ingestion failed for stream kubernetes-metrics: Could not deserialize into JSON object, missing field `schemaUrl`

I think the problem is Parseable’s OTLP/JSON deserialization behavior: it treats the optional schemaUrl JSON field as required. The Collector correctly omits an empty proto3 string during JSON encoding, ref.

Parseable image: quay.io/parseablehq/parseable:v2.9.4

[parmesant]

Looking at the Rust crate for otel-proto -
https://github.com/open-telemetry/opentelemetry-rust/blob/48c5637cafcb8cefd26613f5082a2830a56304af/opentelemetry-proto/src/proto/tonic/opentelemetry.proto.metrics.v1.rs#L64

I guess it is not optional here hence the failure

[parmesant]

https://logg.ing/community
Would love to have a quick chat with you on our Slack 😄

[MatthiasRoels]

I just joined so you can ping me if you want

[DeviousCardi]

I tried to reproduce this and could not, on the exact image reported (quay.io/parseablehq/parseable:v2.9.4, commit 5acfcb6). Posting what I found in case it helps narrow things down.

Ingestion succeeds with schemaUrl omitted

POST /v1/metrics with X-P-Log-Source: otel-metrics, schemaUrl absent at both resource and scope level:

Payload (schemaUrl omitted everywhere)	Result
gauge, scope otelcol/prometheusreceiver	200
sum with exemplars	200
exponentialHistogram	200

schemaUrl is already optional at the serde layer

@parmesant — re: the generated tonic type not being optional, I think the field being String rather than Option<String> isn't sufficient on its own here, because the containing structs carry a container-level #[serde(default)]:

#[cfg_attr(feature = "with-serde", derive(serde::Serialize, serde::Deserialize))]
#[cfg_attr(feature = "with-serde", serde(rename_all = "camelCase"))]
#[cfg_attr(feature = "with-serde", serde(default))]   // <-- makes schemaUrl optional
#[derive(Clone, PartialEq, ::prost::Message)]
pub struct ResourceMetrics { /* ... schema_url: String ... */ }

That attribute is present on every struct that has a schema_url field in the enabled feature set (ResourceMetrics, ScopeMetrics, ResourceLogs, ScopeLogs, ResourceSpans, ScopeSpans), so a missing schemaUrl deserializes to an empty string rather than erroring.

A direct check against the crate, independent of Parseable:

opentelemetry-proto 0.32.0 -- schemaUrl OMITTED:
  [OK] MetricsData
  [OK] LogsData
  [OK] TracesData

This also isn't a version-skew issue: the same #[serde(default)] is present in 0.30.0, and opentelemetry-proto was pinned to 0.32.0 from crates.io in 653c303 (2026-06-17), which predates the v2.9.4 tag (2026-07-11).

What would help

Since the failure is real for @MatthiasRoels but not triggered by a spec-shaped payload, the deciding input is the actual bytes on the wire. Could you capture the exact request body that produced

Could not deserialize into JSON object, missing field `schemaUrl`

for example via a debug exporter or logging exporter on the Collector alongside the OTLP one, or a file exporter? A minimal failing payload would make this a quick fix. It's also worth confirming the running image digest, in case the deployed build differs from the v2.9.4 tag.

Happy to put up a PR once there's a payload that reproduces.