docs: add cluster_name to load assignment entry for static and mostly static

[goz]

Description:
This PR adds missing cluster_name of load_assignment config
for static and mostly-static bootstrap configuration.

Risk Level: Low
Testing: ./docs/build.sh
Docs Changes: Added cluster_name to load_assignment config for static and mostly-static bootstrap configuration.
Release Notes: N/A

Fixes #4056

Signed-off-by: Achmad Gozali gozali@gmail.com

[mattklein123]

@dio it seems a little odd that this field would be required for the static cases when the cluster name is already specified in the config. But indeed it is required so this doc update is correct. Any thoughts here?

[dio]

@mattklein123 it is odd, but since it is also required for the management server to get informed that this field is required is important.

A wild idea will be: since we do validation in bootstrap object level here:

envoy/source/server/server.cc

Line 161 in 0a4bffc

MessageUtil::validate(bootstrap);

we do additional checking and treatment for static resources (i.e. if static, all requirements for cluster name to be filled [cluster_name for cluster.load_assignment equals cluster.name] before the validation).

[mattklein123]

but since it is also required for the management server to get informed that this field is required is important.

Sorry what do you mean? I think it's just an artifact of the message being shared between EDS responses and the local config?

we do additional checking and treatment for static resources (i.e. if static, all requirements for cluster name to be filled [cluster_name for cluster.load_assignment equals cluster.name] before the validation).

Yeah this might be a good thing to do? I will go ahead and merge this since it's required but it seems like we should check they are the same and maybe also do a doc update to say that?

[dio]

Sorry what do you mean? I think it's just an artifact of the message being shared between EDS responses and the local config?

Yeah, I meant this. Sorry for not putting the context that the issue also wants to remove the cluster_name as a required field, but we can't for two reasons.

First, since we also need that to be required for EDS onConfigUpdate validation (but this can be bypassed since we have a check for cluster_name (it has to be the same as the cluster.name) at EDS code)

envoy/source/common/upstream/eds.cc

Lines 61 to 65 in a614808

	// TODO(PiotrSikora): Remove this hack once fixed internally.
	if (!(cluster_load_assignment.cluster_name() == cluster_name_)) {
	throw EnvoyException(fmt::format("Unexpected EDS cluster (expecting {}): {}", cluster_name_,
	cluster_load_assignment.cluster_name()));
	}

Second, I think the only reason why we keep it for now, is to let this validation be generated: https://github.com/envoyproxy/go-control-plane/blob/5f1c666fa3aaa02d95e4fdd95e725f0c5eb7239f/envoy/api/v2/eds.pb.validate.go#L45-L50

[mattklein123]

@dio makes sense. I would probably update the docs to be more clear about how cluster_name is used for load assignment and maybe also put in a check as you pointed out for static clusters that enforces they are the same for clarity?