Redundant flag --reporter.type=grpc when used along with --reporter.grpc.host-port?

[annanay25]

Requirement - what kind of business use case are you trying to solve?

Make agent config flags more intuitive.

Problem - what in Jaeger blocks you from solving the requirement?

Jaeger agent deployment as of now requires the use of --reporter.type=grpc --reporter.grpc.host-port=host:port to report to collector using grpc.

Proposal - what do you suggest to solve the problem or improve the existing situation?

Remove the redundant flag as passing --reporter.grpc.host-port=host:port is sufficient to specify GRPC usage; similar to how --reporter.tchannel.host-port=host:port is sufficient to specify tchannel use.

Any open questions to address

_

[pavolloffay]

similar to how --reporter.tchannel.host-port=host:port is sufficient to specify tchannel use.

this is because of backward compatibility. Before we had only one reporter - tchannel. We will be changing default reporter to grpc in the next release #1218.

[annanay25]

Ah, that makes sense @pavolloffay

We could in the next release, check for the presence of either --reporter.tchannel.host-port or --reporter.grpc.host-port and decide on the reporter.type without having the user explicitly pass the flag.

We could mark this a "good first issue".

[yurishkuro]

I prefer to keep the selector flag until we remove tchannel completely, because if we only rely on the presence of the host-port flag, we have an ambiguity when it's provided for both transports, and we'd need to default to one (or crash, which is perhaps better).

[pavolloffay]

@annanay25 I will close this for now. We also don't want to remove flags as it causes a breaking change so this could happen soonest with the next major version.

[annanay25]

Sure, thanks! :)