2026-06-09
Opt into high-availability Grafana
#kubernetes #grafana #observability #high-availability #rds
Several of you have asked for this: a Grafana that stays up during rolling upgrades, node drains, and AZ disruptions. You can now opt in. When enabled, Grafana runs as two replicas spread across AZs (HPA-scaled up to five), backed by a managed RDS Postgres instance. A pod restart, node drain, or single-AZ outage no longer interrupts the UIfor when you need uninterrupted visibility into your stack.
It’s strictly opt-in: clusters that don’t enable it keep running today’s single-replica setup, with no change in behavior or cost.
Cost
Roughly +$22-26/mo for single-AZ RDS (the default for staging-class clusters), +$37-41/mo for Multi-AZ. The full cost breakdown can be found in our documentation
Already have manual data in Grafana you want to preserve?
If you’ve created dashboards, alerts, users, or annotations through the Grafana UI, ping your Skyscrapers team when you’re ready to flip the flag. We’ll run a one-shot migration job that copies your existing SQLite data into the new RDS database, so nothing is lost. If everything you care about already lives in ConfigMaps, no migration step is needed.
Resources
- Skyscrapers Grafana docs page.
- Design rationale and the alternatives we considered (CNPG in-cluster, Grafana Cloud, AWS Managed Grafana) can be found in our Roadmap item.