vLLM Monitoring and Observability with OpenTelemetry

SigNoz published a guide for monitoring self-hosted vLLM inference servers by sending Prometheus metrics and OpenTelemetry request traces to SigNoz. The guide has operators scrape vLLM's /metrics endpoint on port 8000 every 30 seconds via an OpenTelemetry Collector, dropping the '_created' timestamp series that account for 111 of the 405 exposed series and roughly a quarter of ingested data. Traces are exported directly from vLLM to SigNoz using the OTEL_SERVICE_NAME and OTEL_EXPORT environment variables, bypassing the Collector.

Overview vLLM is an inference server that you run yourself. It serves models over an OpenAI-compatible API. It reports its own health in two ways: Prometheus metrics on an HTTP endpoint, and OpenTelemetry traces for each request. This guide sends both to SigNoz. The metrics answer questions that only the server can answer. How many tokens per second is the GPU producing? How full is the key-value cache, the memory pool that holds attention state for active requests? How many requests wait in the queue? Prerequisites - A vLLM server that you can restart - An OpenTelemetry Collector that can reach the vLLM host. See Install the OpenTelemetry Collector https://signoz.io/docs/opentelemetry-collection-agents/get-started/ - An instance of SigNoz Cloud https://signoz.io/teams/ or Self-Hosted https://signoz.io/docs/install/self-host/ Send vLLM metrics to SigNoz Step 1: Confirm the metrics endpoint vLLM serves Prometheus metrics at /metrics on the same port as the API, and it needs no flag to turn them on. Confirm the endpoint responds: curl -s http://localhost:8000/metrics | head If you changed the port, use that port instead of 8000 . Step 2: Add a scrape job to the Collector Append this scrape job to your existing otel-collector-config.yaml . Do not replace the whole file. otel-collector-config.yaml receivers: prometheus: config: scrape configs: - job name: vllm scrape interval: 30s static configs: - targets: '<vllm-host :8000' metric relabel configs: Drop the Prometheus created timestamp series, which carry no signal. - source labels: name regex: '. created' action: drop Verify these values: - <vllm-host : The hostname or IP address of the machine that runs vLLM. Keep the metric relabel configs block. vLLM exposes a created series next to every counter and histogram, recording when the metric was first observed. On a default server these are 111 of the 405 exposed series. Dropping them removes about a quarter of the ingested data and loses no signal. Then add the receiver to your metrics pipeline: otel-collector-config.yaml service: pipelines: metrics: receivers: prometheus append prometheus to your existing receivers list processors: batch exporters: otlphttp Step 3: Restart the Collector Restart the Collector so that it loads the new scrape job, then watch its logs for scrape errors. sudo systemctl restart otelcol-contrib sudo journalctl -u otelcol-contrib -f docker compose up -d docker compose logs -f kubectl rollout restart deployment/<release -k8s-infra-otel-deployment -n <collector-namespace kubectl logs -f deployment/<release -k8s-infra-otel-deployment -n <collector-namespace The SigNoz K8s Infra chart installs two Collectors. Add the scrape job to the Deployment, which handles cluster-wide scraping, rather than to the per-node <release -k8s-infra-otel-agent DaemonSet. Replace <release with your Helm release name and <collector-namespace with the namespace that runs it. If you run your own Collector under a different name, use that name instead. The MSI installs the Collector as a Windows service named otelcol-contrib , with OpenTelemetry Collector as its display name. Run this in PowerShell: Restart-Service -Name otelcol-contrib Get-EventLog -LogName Application -Source otelcol-contrib -Newest 20 Send a request to vLLM so that the server has something to report. Export Request Traces Optional Metrics tell you how the server behaves as a whole. Traces show where time went inside a single request. Traces go straight from vLLM to SigNoz and do not pass through the Collector, so the scrape job above is unaffected. Set three environment variables, then add one flag to the launch command: export OTEL SERVICE NAME=vllm export OTEL EXPORTER OTLP TRACES PROTOCOL=http/protobuf export OTEL EXPORTER OTLP TRACES HEADERS="signoz-ingestion-key=<your-ingestion-key " vllm serve <your-model \ --host 0.0.0.0 \ --port 8000 \ --otlp-traces-endpoint https://ingest.<region .signoz.cloud:443/v1/traces Verify these values: - <your-model : The model you serve, for example Qwen/Qwen2.5-7B-Instruct . - <region : Your SigNoz Cloud region https://signoz.io/docs/ingestion/signoz-cloud/overview/ endpoint . - <your-ingestion-key : Your SigNoz ingestion key https://signoz.io/docs/ingestion/signoz-cloud/keys/ . You must set the protocol and header variables. vLLM sends no authentication header of its own, and it defaults to gRPC. Restart the server after you change the launch command. Validate Open Metrics Metrics Explorer and filter with service.name = 'vllm' . The Prometheus receiver takes that value from the job name in your scrape job. Metrics appear within one scrape interval. If you enabled traces, open Traces and filter on the value you set for OTEL SERVICE NAME . Each request produces one span named llm request . Metrics Reference | Metric | Type | What it tells you | |---|---|---| | vllm:generation tokens total | Counter | Output tokens produced. Apply a rate to get tokens per second. | | vllm:prompt tokens total | Counter | Prefill tokens processed. | | vllm:prompt tokens cached total | Counter | Prompt tokens served from cache rather than recomputed. | | vllm:time to first token seconds | Histogram | Delay before the first output token. Drives how fast a streaming response feels. | | vllm:inter token latency seconds | Histogram | Delay between output tokens after the first one. | | vllm:e2e request latency seconds | Histogram | Total request duration. | | vllm:request queue time seconds | Histogram | Time a request spent waiting before it ran. | | vllm:num requests running | Gauge | Requests in the current execution batch. | | vllm:num requests waiting | Gauge | Requests waiting to start. A number above zero means the server is saturated. | | vllm:kv cache usage perc | Gauge | Fraction of the key-value cache in use, from 0 to 1. | | vllm:prefix cache queries total | Counter | Prompt tokens looked up in the prefix cache. | | vllm:prefix cache hits total | Counter | Prompt tokens found in the prefix cache. | | vllm:num preemptions total | Counter | Requests the scheduler preempted and re-ran. | Every metric carries model name and engine labels. Group by model name when one server hosts several models. vLLM has no ready-made throughput gauge, so read tokens per second as a rate over vllm:generation tokens total . It also has no prefix cache hit rate metric. Divide vllm:prefix cache hits total by vllm:prefix cache queries total to get one. Metric names vary between vLLM releases. The names above come from v0.29.0. For the full list, see the vLLM metrics reference https://docs.vllm.ai/en/latest/usage/metrics.html . Troubleshooting The metrics endpoint returns 404 Likely cause: the request went to the wrong port, or the server runs behind a proxy that does not forward /metrics . Fix: use the port vLLM serves the API on, which is 8000 by default. Verify: curl -s http://localhost:8000/metrics | head prints lines that start with HELP . The Collector logs a connection refused error Likely cause: vLLM is bound to 127.0.0.1 , so nothing outside the host can reach it. Fix: start the server with --host 0.0.0.0 , and open port 8000 to the Collector host in your firewall. Verify: run curl -s http://<vllm-host :8000/metrics | head from the machine that runs the Collector. Metrics arrive but every value is zero Likely cause: the server has served no requests since it started. Counters and gauges stay at zero until traffic arrives. Fix: send a request to the server. Verify: vllm:prompt tokens total rises in Metrics Explorer. Traces do not appear Likely cause: the server is still exporting over gRPC, which cannot reach SigNoz Cloud. Fix: set OTEL EXPORTER OTLP TRACES PROTOCOL=http/protobuf and OTEL EXPORTER OTLP TRACES HEADERS in the environment of the vLLM process, then restart it. Verify: spans named llm request appear in the Traces explorer. Next steps - vLLM dashboard template https://signoz.io/docs/dashboards/dashboard-templates/vllm-dashboard/ - Set up alerts https://signoz.io/docs/setup-alerts-notification/ on queue depth and time to first token - LLM observability https://signoz.io/docs/llm-observability/ for tracing the application that calls this server Get Help If you need help with the steps in this topic, please reach out to us on SigNoz Community Slack https://signoz.io/slack/ . If you are a SigNoz Cloud user, please use in product chat support located at the bottom right corner of your SigNoz instance or contact us at cloud-support@signoz.io mailto:cloud-support@signoz.io .