FAQ

Short, direct answers to the questions teams ask most before adopting Orbtrace — licensing, data residency, scale, and how it fits next to the tools you already run.

Is there a SaaS version?

No. Orbtrace is on-prem only — that's the entire product positioning. If you can't run it on hardware you control, Orbtrace isn't a fit for your team.

How much does it cost?

One annual license around $20K/year with unlimited hosts. Support (Slack channel, faster SLA) is sold separately by the hour. There's no per-user, per-host, per-GB, or per-custom-metric tax. See the Pricing page for the up-to-date number.

Will my data leave my network?

No, with one exception you control: when you enable RCA / Replay and pick a SaaS LLM provider (Claude, OpenAI, Azure), the prompt contents for that one call go to the provider you chose. The prompt contains span snippets and log excerpts relevant to the incident — exactly what you'd paste into a chat to ask a colleague. You can leave the AI features unconfigured, or run a local Ollama (ORBTRACE_AI_PROVIDER=ollama) so no data ever leaves — air-gapped deployments do exactly that.

What's the smallest machine I can run this on?

Realistically 8 GB RAM for the default stack (it runs three Doris storage nodes); 4 GB only works if you trim Doris to a single node, which the shipped Compose file doesn't do out of the box. For real production traffic plan for 16–32 GB RAM and SSD storage. See Installation → requirements.

Can I migrate from Datadog / Grafana + Tempo + Loki / SigNoz / HyperDX?

Yes. Migration is mechanical because everyone speaks OpenTelemetry. Point your existing OTel SDK config at Orbtrace's Collector and the data flows in. Dashboards have to be rebuilt; alert rules can be exported as YAML and rewritten against Orbtrace's syntax. Premium support includes an opinionated 2-4 week migration engagement.

Does it work in an air-gapped environment?

Yes. Nivorbit builds an orbtrace-<version>-airgapped.tar.gz on demand (it bakes in every image, so it's too large for a release asset) and hands it over out-of-band. You copy it across, load the images with bash images/LOAD.sh, and install offline via Ansible or docker-compose, running a local Ollama for the AI features. The license is an offline signed token verified against a key baked into the product — no phone-home. See Air-gapped operation and Licensing.

Which LLM is the default and why?

Anthropic Claude is the default because it has the strongest "cite your sources" behaviour in our benchmarks — RCA's grounded-citation pattern works best with it. OpenAI and Azure OpenAI are equally capable for incident summarisation; Ollama models are weaker but acceptable for medium-complexity incidents and useful when no data can leave.

Can I run two providers at once?

No. One Orbtrace instance, one active provider. The reason is consistency: RCA outputs from different models cite spans in different shapes, and mixing them makes incident history hard to compare. Switching providers is hot (no downtime), but past RCAs are not retroactively re-run on the new provider.

How is this different from Grafana?

Grafana is a dashboarding UI you plug into multiple backends (Prometheus for metrics, Loki for logs, Tempo for traces). Orbtrace is the whole stack — backend, dashboards, alerting, RCA, replay. If you love Grafana's dashboards but hate gluing five tools together, you're our customer.

How is this different from Tempo / Jaeger?

Tempo and Jaeger only handle traces. Orbtrace handles logs, traces, and metrics in one query language, with Causal RCA and Time-Travel Replay on top. They're not really competitors — they're a slice of what Orbtrace does.

How is this different from HyperDX?

HyperDX was a data-exploration product (now part of ClickHouse Inc.). The interaction model is similar; the differentiation is that Orbtrace is on-prem-only, includes the RCA + Replay engines for diagnosis (not just exploration), and bundles a full alerting platform.

Does this scale?

A single Doris cluster comfortably ingests 100K+ events/sec on modest hardware. The OpenTelemetry Collector you run in front scales horizontally (it's standard contrib — add replicas behind a load balancer). The Spring Boot backend is stateless and scales by adding pods. The bottleneck in practice is Doris BE memory and disk I/O — we ship a sizing guide that turns "X events/sec and Y days of retention" into a hardware recommendation.

What happens when my license expires?

It depends on the license type, and it's deliberately gentle — losing observability mid-incident would be the worst failure mode. A perpetual license never stops (the date is only your support window). A subscription steps down gracefully: a warning banner, then a grace period at full function, then a degraded state where ingestion and existing alerts keep running but the interactive surface (ad-hoc search, on-demand AI, creating/editing alerts/SLOs/dashboards) is paused until you renew. A trial locks to the license screen at expiry. Your data is never deleted. Full detail in Licensing.

Is there a free trial?

Yes — a 30-day pilot license (extendable on request). Email sales@nivorbit.com with your hardware target.

What about Kubernetes?

There's a Helm chart. See Deploying on Kubernetes.

What about Windows servers?

Not supported. The product is Linux-native (Doris and the Spring Boot server). If your team only runs Windows, use the Kubernetes path on a Linux node pool.

How do I report a bug?

Standard support: a GitHub Issue on nivorbit/orbtrace-issues (a public-issue tracker, not the source repo). Premium support: your private Slack channel.

Where do I see what's new in each release?

The Orbtrace landing site's Changelog page, plus a "What's new" toast inside the app on each version bump.

Is the source code open?

No. Orbtrace is closed-source commercial software. (We don't ship any custom OpenTelemetry Collector component — you run the stock upstream otel/opentelemetry-collector-contrib, which is open-source and not ours.) The Orbtrace product itself is not open-source.

Can I sponsor a feature?

Yes. Premium support customers get a "feature voting" channel; sponsored features are scoped and quoted by our team. Email sales@nivorbit.com.