Ceph » Orchestrator

In the current design ceph can run multiple instances of mgr. The mgr is a core component of ceph that runs important modules such as dashboard or cephadm. Only one mgr can be active and it's the one that acts as access point for the different services provided by the mgr. Right now for example dashboard running in the standby just performs a redirection to the active mgr. From the outside, Prometheus for example has to know what's the IP of the active mgr in order to scrape metrics or to get the dynamic service discovery information. In case of a mgr fail-over Prometheus config must be updated to point to the new one. The idea of this feature is to add HA support for the mgr by using the ingress service. This way mgr will run behind a stable VIP and users should not care anymore to know what's the IP of the current active instance.