Q&A Source Host Metrics in OTEL format

[ashinde-ai]

Question

Vector's host_metrics source currently does not natively support exporting metrics in OpenTelemetry Protocol (OTLP) format, despite Vector having an opentelemetry sink. This creates a significant gap for users wanting to integrate Vector's excellent host metrics collection with OpenTelemetry-native observability platforms like SigNoz, Jaeger, and others.

Current Situation

• ✅ Vector's host_metrics source collects comprehensive system metrics (CPU, memory, disk, network, filesystem, load)
• ✅ Vector has an opentelemetry sink that supports OTLP format
• ❌ The host_metrics source produces Vector's internal metric format, not OTLP-compatible data
• ❌ Direct connection between host_metrics → opentelemetry sink fails with data type mismatches

Vector Config

data_dir: "/tmp"

Disable acknowledgements

acknowledgements:
enabled: false

Health checks

healthchecks:
enabled: true
require_healthy: true

API for monitoring

api:
enabled: true
address: "127.0.0.1:8686"
playground: false

Sources

sources:
host_metrics:
type: host_metrics
scrape_interval_secs: 30
collectors:
- cpu
- disk
- load
- memory
- network
- filesystem
- host

vector_metrics:
type: internal_metrics

No transforms - using native OpenTelemetry sink

Sinks

sinks:

Direct to SigNoz using native OpenTelemetry sink

signoz_direct:
type: opentelemetry
inputs:
- host_metrics
protocol:
type: http
uri: "https://ingest.us.signoz.cloud:443/v1/metrics"
encoding:
codec: json
headers:
signoz-ingestion-key: "****************************"
compression: gzip
batch:
max_events: 100
timeout_secs: 10

Debug console output to see raw metrics

debug_console:
type: console
inputs:
- host_metrics
encoding:
codec: json

Prometheus exporter for monitoring Vector itself

prometheus_exporter:
type: prometheus_exporter
address: "127.0.0.1:8688"
inputs:
- vector_metrics

Vector Logs

Host Metrics not going OTLP format

{"name":"cpu_seconds_total","namespace":"host","tags":{"collector":"cpu","cpu":"0","host":"signoz","mode":"idle"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":1984290.42}}
{"name":"cpu_seconds_total","namespace":"host","tags":{"collector":"cpu","cpu":"0","host":"signoz","mode":"io_wait"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":475.85}}
{"name":"cpu_seconds_total","namespace":"host","tags":{"collector":"cpu","cpu":"0","host":"signoz","mode":"nice"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":38.08}}
{"name":"cpu_seconds_total","namespace":"host","tags":{"collector":"cpu","cpu":"0","host":"signoz","mode":"system"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":5672.56}}
{"name":"cpu_seconds_total","namespace":"host","tags":{"collector":"cpu","cpu":"0","host":"signoz","mode":"user"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":8602.07}}
{"name":"cpu_seconds_total","namespace":"host","tags":{"collector":"cpu","cpu":"1","host":"signoz","mode":"idle"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":1985274.35}}
{"name":"cpu_seconds_total","namespace":"host","tags":{"collector":"cpu","cpu":"1","host":"signoz","mode":"io_wait"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":616.69}}

{"name":"network_transmit_errs_total","namespace":"host","tags":{"collector":"network","device":"eth0","host":"signoz"},"timestamp":"2025-11-05T10:40:18.798559298Z","kind":"absolute","counter":{"value":0.0}}
2025-11-05T10:40:20.138751Z ERROR sink{component_kind="sink" component_id=signoz_direct component_type=opentelemetry}:request{request_id=1}: vector::sinks::util::retries: Not retriable; dropping the request. reason="Http status: 400 Bad Request" internal_log_rate_limit=true
2025-11-05T10:40:20.140356Z ERROR sink{component_kind="sink" component_id=signoz_direct component_type=opentelemetry}:request{request_id=1}: vector_common::internal_event::service: Service call failed. No retries or retries exhausted. error=None request_id=1 error_type="request_failed" stage="sending" internal_log_rate_limit=true
2025-11-05T10:40:20.140427Z ERROR sink{component_kind="sink" component_id=signoz_direct component_type=opentelemetry}:request{request_id=1}: vector_common::internal_event::component_events_dropped: Events dropped intentional=false count=545 reason="Service call failed. No retries or retries exhausted." internal_log_rate_limit=true

[pront]

Hi @ashinde-ai, this is a well documented limitation. I recommended creating a feature request with a proposal for this. Without deep diving into this, I can imagine a VRL function that transforms Vector metric events to OTLP events or baking this in the otlp encoder. In any case, it's worth a discussion and maybe it can inspire an external contributor to implement this.

[pront]

As a workaround, which might be a rabbit hole, you can use do a metric-to-log conversion followed by a complex remap transform to mutate your events.