Share your GreptimeDB story

[killme2008]

We'd love to hear how you're using GreptimeDB in the wild.
Some things we're curious about:

• What are you monitoring or storing?
• What did you migrate from (Prometheus, Loki, ES, ClickHouse, etc.)?
• What worked well? What didn't?
• Any performance numbers or cost savings you can share?

No format required — a few sentences is great, a detailed write-up is even better. We may feature some stories (with your permission) in our docs or blog to help other users evaluating GreptimeDB.
Thanks for being part of this community 🙏

If you'd prefer not to share publicly, feel free to DM me(dennis zhuang) on Slack — happy to chat privately.

[heiwen]

We're using GreptimeDB as the observability backend for Hebo Platform, an AI gateway. We store OpenTelemetry traces from AI workloads — token usage, inference latency, model metadata, tool invocations — using the Gen-AI semantic conventions.

Migrated from a planned ClickHouse setup before we had committed to it. The native OTLP ingestion was the deciding factor: it let us go directly from SDK to queryable SQL without a queue/worker/transformation pipeline in between.

What worked well: the disaggregated storage model (object store as primary storage, not a cold tier), the dynamic schema for span attributes, and the overall simplicity of the setup — single Docker container locally, Helm for production, same config throughout.

What didn't: we hit a few bugs along the way — Unicode encoding in query results, binary-encoded timestamp parameters hanging in cluster mode, S3 auth via EKS Pod Identity. The team was fast to attend to all of them, which gave us confidence to keep going.

We wrote up the full reasoning here if useful: Why we bet on GreptimeDB instead of ClickHouse

[killme2008]

Thanks for sharing, @heiwen.It’s a great story.

[killme2008]

Personal project: https://github.com/tma1-ai/tma1

Local-first observability for AI agents with a built-in dashboard. Track costs, latency, and anomalous behavior.

Built on GreptimeDB, using OpenTelemetry and cc/codex session hooks, plus the GreptimeDB flow engine to generate metrics.