3 ms·
What always seems to be missing from these things is gRPC monitoring - there's a defined HealthCheck endpoint, even, specifically for this situation. Maybe peo
by zimpenfish 6y ago
What always seems to be missing from these things is gRPC monitoring - there's a defined HealthCheck endpoint, even, specifically for this situation. Maybe people just don't run gRPC services as much as the hype would imply?
- enz 6y agoWhy would this be so valuable compared to basic HTTP health check?
- zimpenfish 6y agoIf you've got a gRPC API, you don't necessarily have anything serving HTTP that would correlate to that service - I know we don't, for example; the only HTTP is for serving static files, everything else is gRPC, and HTTP checks wouldn't spot any problems.
- capableweb 6y agoDon't people usually use gRPC over HTTP?
- zimpenfish 6y agoThe core protocol uses HTTP/2 as a transport, yes, but you'd have to convert the conceptual gRPC request to the protobuf-encoded version and somehow pass that to the HTTP/2 endpoint as a binary blob, then decode/match the binary blob you get back - which none of the status pingers really offer either. (You could use `gRPC-web` to make JSON-encoded requests but you need specific support on the server end for that.)