xds: require EDS service name in new-style CDS clusters (gRFC A47) #6438
Conversation
…tp' resource name (gRFC A47)
| { | ||
| name: "xdstp cluster resource with unset EDS service name", | ||
| resource: testutils.MarshalAny(&v3clusterpb.Cluster{ | ||
| Name: "xdstp:foo", |
There was a problem hiding this comment.
Should this be xdstp:///foo? I understand it doesn't matter for the purpose of this unit test. But was just curious if xdstp:foo was a valid name at all.
There was a problem hiding this comment.
The two are equivalent under RFC 3986 parsing.
Also I took it from the test case in C without really thinking too hard about it. 😆
I'm not sure what does or should happen in this case, though. Does/should it just not work? Does/should it use the channel's default authority? Does/should it use the authority for the CDS resource?
The only meaningful thing this cluster.name field seems to do for us is this:
Looks like we'd just ignore it (or use the default authority?) to determine the load reporting server. I'm not sure if that's right or not, TBH - do you have any intuition about it?
This PR implements the changes described by grpc/proposal#377
RELEASE NOTES: