multiple prometheus

[sstarcher]

Would it be possible to support multiple prometheus from a single Pyrra? We run the Prometheus operator and create multiple servers in different namespaces.

[sstarcher]

Either that or the ability to separate the UI and the operator. Where a user could deploy the UI to a Kubernetes cluster without Pyrra and deploy the SLO operator to a cluster with multiple Prometheus.

[metalmatze]

We run Pyrra at Polar Signals against two Prometheus replicas operated by the Prometheus Operator via kube-prometheus. What exact problems are you seeing that you want to solve? For us, this works fine.
The easiest thing to solve deduplication would probably be running the queries through a Thanos Querier in front of the two Prometheus replicas.

[sstarcher]

Not replicas, but entirely separate prometheus servers with separate metrics. We could certainly run Thanos and rollup metrics, but we specifically don't and don't really want the metrics to comingle.

Currently we could run multiple Pyrras, but it would be better for us if the UI could pull from multiple prometheus servers which is what Grafana does do for us.

[metalmatze]

Got it. I can see how that might makes sense but is not on the roadmap.
For now, you'll have to make the data accessible through the same API or run two Pyrra instances.