We tested the on-prem central monitoring using Kafka/Grafana and it seemed to work great with the default dashboards included. The documentation in docs.memsql is very limited and can be a bit confounding with the terminology. MONITORED_CLUSTER (a cluster being monitored) and COLLECTING_CLUSTER (Central monitor which runs kafka and grafana datashboards) would avoid the confusion. But overall a great win for memsql… if we can get 2 dozen pre-canned dashboards (with json) that would be fabulous… we are already starting to implement it ubiquitously.