PoC for operating the k8sclusterreceiver

[a-thaler]

Description

In the current architeture the telemetry module operates a metric gateway for enrichment and filtering as well as an agent for collecting node-affine data. The k8sclusterreceiver is retrieving data from the k8s apiserver only and with that uses a pattern which is not yet covered in the architecture.
Goal is to come up with a proposal on how to operate this receiver to have it available as another input source. Data duplication should be avoided but downtimes preventing as much as possible (a node crash should not cause a long downtime of the functionality).

Alternatives are mainly to run it on the agent, the gateway or in a new component (latter should be avoided to reduce complexity and resource usage). When running in multiple instances only one instance is active in best case, the rest on stand-by.

Criterias

• Provide a concrete concept on how to operate the receiver
• Proof that the concept will solve the problem
• The proposed solution should try to avoid data duplication (to keep resource usage low)
• The proposed solution should achieve a close to HA-setup (downtimes of seconds is fine)

Reasons

Attachments

Release Notes

[skhalash]

PoC: https://github.com/skhalash/leaderreceivercreator