grpc/grpc

gRPC gitter chat

Peter Sumner

@saltxwater

Hello! First time poster here. Just a question regarding DNS and connecting from (C#) clients: We're registering all our (docker) services in consul and they have random ports assigned by nomad. Our current workflow involves querying consul for the service info (address and port) and passing those straight through to "new Channel(host, port, ...)"
We've now got consul set up as a dns resolver so I would LIKE to just call 'new Channel("dns://service.consul.mycompany.com")' and have the port determined by the DNS response (SRV records which include the port).
Is this possible with the current state of things? If not, do you think it is a reasonable feature to request on github issues?

Richard Belleville

@gnossen

@matrixbot grpcio is the best supported gRPC Python implementation. If you're having performance issues, that is something that we'd like to know about. Can you share any more of the particulars of your setup? Perhaps even code samples? What sorts of protos are you using? What are you doing inside of your servicers?

Richard Belleville

@gnossen

@matrixbot It'd be helpful if you could gather that information together into an issue at https://github.com/grpc/grpc/issues/new/choose

matrixbot

@matrixbot

xionbox Richard Belleville (Gitter): okay, I'll see what I can provide (it's a closed source project). Note that the timing numbers I gave earlier were actually based on an incorrect timing method. I've fixed this now, and re-running the benchmark

Luca Lolli

@dashandslash

@ejona86 unfortunately there is no c++ api to shutdown a channel or explicitly change its state to GRPC_CHANNEL_SHUTDOWN

Daniel

@lumpidu

@dashandslash : grpc::CreateChannel returns a shared_ptr, just call reset() on it

Luca Lolli

@dashandslash

@lumpidu That will reset the shared_ptr destruct the object, not in any safe way...if you have any "wait" on the channel....well you might know what it´s going to happen..
What I am trying to find out is how to change the GRPC_CHANNEL_READY state to GRPC_CHANNEL_SHUTDOWN

Daniel

@lumpidu

@dashandslash : https://github.com/grpc/grpc/blob/master/src/cpp/client/channel_cc.cc: it will call grpc_channel_destroy(), which according to the documentation, "close and destroy a grpc channel". Why don't you think it's safe ?

x0a1b

@x0a1b

Is there a way specify maximum number of requests per a server should handle per channel? This will help client's keep cycling connections and avoid traffic skew as pods in our cluster come up or go down.

Eric Anderson

@ejona86

@x0a1b, no. I assume you are using pick_first (the default) load balancer on the channel. Proper thing to do is enable max connection age on server-side. That puts the service in control of its configuration.

What language is your server in?

Luca Lolli

@dashandslash

@lumpidu the reason why it´s not safe to destroy the shared_ptr is that if the channel is executing aWaitForStateChange on a thread...this will crash.
Basically I want to do a WaitForStateChange(GRPC_CHANNEL_READY, timeout) in a separate thread, and somehow explicitly change the state of the channel or call a shutdown api to break the WaitForStateChange

Yash Tibrewal

@yashykt

@dashandslash We don't have an API to explicitly change the state of the channel from READY. Waiting for the timeout to expire would be the most straightforward option.

x0a1b

@x0a1b

@ejona86 server is in Java and I think maxConnectionAge will help

cranemogul

@cranemogul

Hi, I have a question regarding protoc plugin development in Go.

I'm processing service structures and would like to read method options; how can I do that?

from MethodDescriptorProto I can only get Options, but what useful data can I get from them?

Eric Anderson

@ejona86

You want the ServiceDescriptorProto which has method descriptors inside

Wait...

Now I'm confused. What do you mean by "only get Options". What additional data do you expect?

cranemogul

@cranemogul

well, I can't get anything useful from Options object. For example, UninterpretedOption object is nil. In the file1.proto file, which I process, the method looks like
rpc M1(MI) returns (MO) {
option (google.api.http) = {
post: "/v1/m"
body: "*"
};
}

Eric Anderson

@ejona86

I think there are proto2 extensions in it.

Each option is a proto2 extension.

cranemogul

@cranemogul

how can I extract this data - for example, "/v1/m" - from Options?

ok, then how should I proceed? I've tried GetExtension, HasExtension... compiler says there are no such things in MethodOptions

Eric Anderson

@ejona86

So this is that option. https://github.com/googleapis/googleapis/blob/master/google/api/annotations.proto#L30

So you want the HttpRule extension from the MethodOptions message.

cranemogul

@cranemogul

yes

Eric Anderson

@ejona86

Looks like it is a function (not a method) you have to call
https://github.com/golang/protobuf/blob/master/proto/extensions_test.go#L54

cranemogul

@cranemogul

am I missing some import?..

Eric Anderson

@ejona86

A bit strange, but works I guess.

cranemogul

@cranemogul

so I should mention HttpRule structure as a parameter to GetExtensions?..

Eric Anderson

@ejona86

I think there should be a ExtensionDesc in the generated code for annotations.proto. You would pass that to GetExtensions()

Looks like the name will be prefixed with E_

https://stackoverflow.com/questions/28815214/how-to-set-get-protobufs-extension-field-in-go

cranemogul

@cranemogul

I saw this, but not sure how to translate that into Go

cranemogul

@cranemogul

GetExtensions() failed: proto: not an extendable proto.Message

    exts, err := proto.GetExtensions(curMethod, []*proto.ExtensionDesc {
        pb.E_Ext_More,
        pb.E_Ext_Text,
        //HttpRule http = 72295728;
    })
    if err != nil {
        log.Fatalf("GetExtensions() failed: %s", err)
    }

Eric Anderson

@ejona86

I'm not sure.

cranemogul

@cranemogul

it's good it compiles at all :) - sometimes it's not easy to achieve... structures are unclear...

here I had to import pb "github.com/golang/protobuf/proto/test_proto"

thank you Eric. Will think some more...

asaf david

@asafdav2

Hi, a question regarding health check - my service is running in GCP, and needs to respond to GCP healthchecks which are one of TCP, HTTP, H2 but not gRPC. is there any way to make it talk with a HC service that implements the gRPC HC protocol? I have envoy sidecar if that's helpful (I see that envoy has ability to HC gRPC cluster but i'm not sure if i can leverage it)

Eric Anderson

@ejona86

@asafdav2, TCP health check would be normal today. We've looked into gRPC-native health checks, but I don't think there is anything to share at the moment.

Alexey Edelev

@semlanik

Hi I sent request using this form https://docs.google.com/forms/d/119zb79XRovQYafE9XKjz9sstwynCWcMpoJwHgZJvK74/viewform?edit_requested=true to add my project to gRPC community, should I expect any response?

Eric Anderson

@ejona86

@semlanik, I don't really know, but I'd assume so given the many projects added to grpc-ecosystem.

Pablo Cariel

@pcariel

Hi, vertx hazelcast using eventbus ?

Dmitrii Ustiugov

@ustiugov

hi! I am trying to build gRPC with cmake from the external project
grpc/examples/cpp/helloworld/cmake_externalproject

The issue is that I cannot configure the Release build to enforce compiler's "-O2" optimizations without hacking through all dependencies' CMakeLists.txt (e.g., gRPC, protobuf, zlib, etc)

It's difficult to believe that there is no easy way to switch the optimizations on and off in gRPC....

cranemogul

@cranemogul

Is there a way to generate Swagger files (using Swagger protoc plugin) only for services, and not generate Swagger files for other .proto files, which don't contain services?

Where communities thrive

People

Repo info

Activity