I've been analyzing the new Grafana Cloud pricing model, specifically the "Flexible" plans for logs, and I'm seeing some serious cost traps for anyone with high-volume, variable workloads.
The shift from a simple per-gigabyte ingested model to a combination of **Active Series hours** and **Query hours** is a classic move towards hidden complexity. You now pay for:
* **Log Active Series hours:** Based on unique label sets, not just volume ingested. High-cardinality environments will get obliterated.
* **Log Query hours:** Every query, automated or user-driven, adds to the bill. This makes dashboards and alerts a direct cost center.
* **Ingested data:** Still there, but now just one of three levers.
The problem is the coupling. A spike in log volume increases series, which in turn can make your queries more expensive to run. It's a multi-dimensional pricing model that's very difficult to forecast.
Has anyone run a direct cost comparison between the old and new model for a real workload? I'm particularly skeptical about the break-even points they suggest. If you have 100+ GB/day of logs with moderate cardinality, I suspect the new "Flexible" plan could be 2-3x more expensive than the old Pro plan unless you aggressively optimize series count.
What's your experience? Have you been forced to implement drastic cardinality limits or query throttling to keep costs manageable?
cost optimization, not cost cutting