QDRANT CLOUD OBSERVABILITY

See Exactly What Your Cluster Is Doing

Qdrant Cloud exposes query latency, throughput, and infrastructure metrics in OpenMetrics format. Point Prometheus, Grafana, Datadog, or another observability tool at these endpoints to track performance against your RPS targets and catch degradation early, or view baseline CPU, memory, and disk usage directly in the Qdrant Cloud Console.

Start Free Explore the Docs

WHY IT MATTERS

Tune latency and throughput together.

Watch request latency histograms and requests per second side by side, then push throughput until you find the real ceiling for your workload.

Iterate on quantization with a clear signal.

Latency and resource metrics update as you adjust quantization settings, so each experiment gives you a measurable before and after.

Compare workloads on one metrics view.

Track request volume, pending operations, and hardware usage per collection, so you can see which workload is consuming resources and allocate capacity deliberately.

WHAT YOU CAN SEE

Full Visibility Into Your Cluster

Per-collection visibility.

If your collection uses multi-vector representations for late-interaction retrieval, Point counts, vector counts, pending optimizations, and per-collection hardware metrics are labeled by collection, so a spike can be traced back to the workload causing it, including on clusters running many tenants or workloads.

Check out Alerting

Query latency and throughput signals

Request duration histograms, averages, min/max, and per-service load balancer timings, plus total and failed request counters for tracking RPS and error rates.

Built-in metrics in the Console

CPU, memory, and disk usage are available in the Qdrant Cloud Console.

Alerting that's already on

Qdrant Cloud monitors cluster health for you on Managed Cloud and Hybrid Cloud, memory, disk, node status, and CPU balance, and emails you automatically when something needs attention.

Observability That Fits Your Deployment

Qdrant Cloud surfaces infrastructure and request-level metrics.

Application-layer tracing, LLM call chains, and end-to-end retrieval quality scoring sit above the database layer and require a separate tool connected to your application code. If you are sizing a complex multi-tenant observability setup or need custom alerting rules validated against your specific workload, contact Qdrant to discuss your architecture.

Talk Through Your Architecture

Managed Cloud

Hybrid Cloud

Private Cloud

Open Source

Works with your observability stack (Prometheus, Grafana, Datadog, any OpenMetrics tool)

Metrics at a glance in the Cloud Console

N/A

Automatic alerting on cluster health

Your own Alerts rules, in your own stack

Your own rules, in your own stack

Where your telemetry goes

Qdrant operates the infrastructure

Your stack & Qdrant's platform, both

Your stack only, air-gapped by design

Your stack only

FAQs

What metrics does Qdrant Cloud expose?

Qdrant Cloud coverage includes query latency histograms, request counters (RPS and error rates), memory and CPU usage, and per-collection request, hardware, and configuration metrics. Learn more: https://qdrant.tech/documentation/cloud/cluster-monitoring/

Do I have to build my own dashboards?

No. Qdrant ships a pre-built Grafana dashboard as importable JSON with built-in views and graphs for monitoring your clusters. Import it, then customize as needed. Get the dashboard: https://github.com/qdrant/qdrant-cloud-grafana-dashboard

Can I use my existing Prometheus and Grafana setup?

Yes. Point your Prometheus instance at the metrics endpoints using a read-only API key (supported as a Bearer token), and import the pre-built dashboard into Grafana. The docs include ready-to-use ScrapeConfig examples for Managed Cloud and ServiceMonitor examples for Hybrid Cloud. Learn more: https://qdrant.tech/documentation/ops-monitoring/managed-cloud-prometheus/

Does Qdrant integrate with Datadog?

Yes. Configure the Datadog Agent's OpenMetrics check to scrape Qdrant's endpoints; the documentation includes a worked Autodiscovery configuration. Any other platform that ingests Prometheus/OpenMetrics data connects the same way.

How does observability work on Hybrid Cloud versus Managed Cloud?

On Managed Cloud, Hybrid Cloud, and Private Cloud expose different levels of infrastructure detail. Learn more: https://qdrant.tech/documentation/ops-monitoring/hybrid-cloud-prometheus/

Can I get per-collection or per-tenant metrics?

Qdrant Cloud surfaces metrics at the collection level: request counts, pending operations, hardware usage, and segment statistics per collection. That lets you set alerting thresholds that reflect each workload's traffic and resource profile. Latency metrics are reported at the node level, and metrics aren't labeled by tenant ID within a shared collection, so per-tenant latency tracking happens at the application layer.

What alerting options are available?

Qdrant exposes the raw metric signals; we also have alerting rules in the UI and via email that you can use, or you configure alerting rules in your own stack, whether that's Prometheus Alertmanager, Datadog monitors, or the incident tooling already wired into your Grafana. For help validating alerting thresholds against your specific workload, contact Qdrant. https://qdrant.tech/documentation/cloud/cluster-monitoring/#alerts

Does Qdrant see my vectors or collection data when it collects telemetry?

No. Your vectors, payloads, and queries stay inside your cluster on every deployment mode. What Qdrant's telemetry collects is infrastructure-level: on Managed Cloud and Hybrid Cloud, that means metrics like CPU, memory, and disk, plus cluster metadata such as names, labels, and collection counts. It never includes the contents of your database. Storage volumes are encrypted at rest, and API keys are stored as hashes.

On Hybrid Cloud and Private Cloud, the isolation goes further: your database, stored data, API keys, backups, and cluster logs all stay on your own infrastructure with no Qdrant access. On Private Cloud, which is air-gapped, Qdrant sees no infrastructure metrics or metadata either.

Learn more: https://qdrant.tech/documentation/cloud-security/