apolcyn
apolcyn
Even if we did change the way that GCE detection is done, there are already a bunch of clients in the wild (e.g. from older gRPC releases) that depend on...
^ +1 I can't speak for everywhere, but my intended usage is not particularly performance concerned with the writev call.
In the core library, when someone composes a call credentials with an insecure channel creds, the call creds are just silently dropped. The motivation for that was security: call creds...
Even though it's not really breaking anything for `Grpc.Core` (since the headers still won't go out), it's a regression to go from obvious error to silent error (not sending headers)....
> @apolcyn, does it seem wrong that initial RPCs fail if gRPCLB server is unavailable, when fallback is available but hasn't been tried yet? It appears later RPCs would succeeded...
We do have a preliminary sketch for this in the Google groups chat btw, but it is looking like it will take some more time than initially thought.
cc @ctiller please merge when you have a chance
Sorry for the delay @johnnyshields. Aiming to go through this this week.
Thanks. I probably won't get a chance to take a good look at this soon (it's been a while). But with a quick look I believe that indeed this does...
@easwars or @dfawley can you merge? I don't have merge access