I am planning an HA setup in a highly automated environment. All Grafana configuration is provided upon deployment of Grafana, fully generated. Grafana will be deployed HA in a Kubernetes cluster.
Sessions we will put in a shared, separate session store (Redis).
My question: in such a scenario, can we avoid using a shared MySQL for the Grafana instances and just stick to the SQLlite storage? There will not be any operational data besides the sessions, is my understanding, correct?
As we will use alerting feature, what about the dedupe logic? Is that using the shared storage - making a shared DB a requirement?