[Fleet] Fleet uses the wrong configuration key for agent.download.sourceURI

[cmacknz]

Fleet is creating agent policies with agent.download.source_uri when the agent expects it to be agent.download.sourceURI. This has caused several bugs with agent artifact downloads not working correctly, causing the agent to have to handle both formats. See elastic/elastic-agent#1252 for one example.

The agent configuration entry for the download URI is defined here:

	// SourceURI: source of the artifacts, e.g https://artifacts.elastic.co/downloads/
	SourceURI string `json:"sourceURI" config:"sourceURI"`

I am not entirely sure where Fleet defines this, but one reference to source_uri instead of sourceURI can be found at:

kibana/x-pack/plugins/fleet/common/types/models/agent_policy.ts

Lines 104 to 113 in e8e1203

	agent?: {
	monitoring: {
	namespace?: string;
	use_output?: string;
	enabled: boolean;
	metrics: boolean;
	logs: boolean;
	};
	download: { source_uri: string };
	};

Fixing this would require migrating existing agent policies to the expected configuration key.

I am less interested in fixing this particular instance of Fleet being out of sync with the Agent, and more interested in defining a way to prevent this type of problem in the future.

[elasticmachine]

Pinging @elastic/fleet (Team:Fleet)

[nchaulet]

@cmacknz is there somewhere a JSON schema of the agent policy that we can maybe use in our test to validate that we always generate a correct agent policy.
I am not sure what will be the best way to share this but if this schema live in the elastic agent repo we could have some automated PR that update kibana with it, (like the one for kubernetes agent manifest)

[cmacknz]

Thanks, there isn't a JSON schema for the agent policy unfortunately. Creating something like this has come up a few times, I'll look into creating one.