Observer Plugin
The main function of observer is to watch observabilityIndicant, to obtain metrics data periodically, and the plugin to be implemented is to serve as a data provider.
To implement a plugin, you only need to implement a GRPC interface. The interface only needs to contain three functions:
GetPluginName
This interface is mainly used to get the plugin name, so that the observer can ignore some crs that it does not pay attention to.
GetCapabilities
Get the capabilities supported by the current plugin, for example, get the maximum value of CPU over a period of time, or the average value of memory, etc.
When the customer writes the aggregated value of the indicator to be obtained in cr, the observer will integrate the support ability obtained from the current plugin with what the user writes. For example:
The user needs the aggregation of max and min with the cpu, but the plugin only provides the aggregation capability of min, so when the observer plugin requests data, it will only request the aggregation value of min.
GetMetrixs
Returns aggregated metrics data for the target Pod.
After implementing the relevant interface, start the rpc process and deploy it together with the observer program (sidecar mode), the two communicate through the unix domain socket.