channelBuilder.enableRetries() then it enables the subsystem for the entire channel. That is the most likely culprit. Try removing that enableRetries() call and see if the issue disappears. If it does "resolve" the problem, then it is clearly a bug in the retry code, so please file an issue for us to track down the bug.
anybody has an example of a custom grpc auth-service preferably in python and how it sits in the k8s for service to service communication e.g. i want a separate auth-service deployed which generates e.g. JWT token by checking username/password and then a service (or interceptor, im not sure) that is in front of every service in cluster maybe that validates the JWT token
1 reply
for gRPC(Ruby) can someone point me to a method to add details to a gRPC exception.
Something similar to https://pkg.go.dev/google.golang.org/grpc/internal/status#Status.WithDetails
I think I found the solution
field_error = Base::FormErrors::Field.new(name: 'name', error: 'must be unique')
form_errors = Base::FormErrors.new(fields: [field_error])
raise GRPC::InvalidArgument.new('Validation Failed', 'grpc-status-details-bin' => form_errors.to_proto)I have to use the metadata to pass the details
1 reply
Hi, Im new to gRPC and have a couple questions that I got from the project I was assigned:
- I have this case on MongoDB:
@Document(collection = "Notebook") data class NotebookModel{
@id val id: String,
val name: String,
val desc:String,
val noles:List<Notes>,
val owner:Owner
}
the problem is that Owner is a proto generated class, so when I try to list or find elements, Im getting an error:
"java.lang.IllegalStateException: Cannot set property fields because no setter, no wither and it's not part of the persistence constructor private com.google.protobuf.UnknownFieldSet()!"
I just want to know if this error is because the class only way to instantiate is with a builder and when Mongo is trying to set the
object values , it can't
2 replies
- my second question is an odd behavior Im having when Im building a response:
I have a simmilar code to this one:
SaveNoteBookResponse{
EnumStatus status: 1 ,
String Message: 2
}
func saveNotebook():SaveNoteBookResponse{
var status:EnumStatus = Status.inprocess
var message = "in proccess"
try{ var status= Status.Success
var message = "Success!"
}catch(){
var status = Status.Error
var message = "Ups something went wrong!"
}
val response = SaveNoteBookResponse.newBuilder().setStatus(status).setMessage(message).build()
return message}
for some reason, when I set Status, is not beeing set on the response, instead of having:
{
"status": "Success" ,
"Message": "Success!"
}
Im getting :
{
"Message": "Success!"
}
The same happens when there is an error , the only way I made it work is setting the enum directly when Im building the response:
val response = SaveNoteBookResponse.newBuilder().setStatus(status.Success).setMessage("message").build()
Is there a restriction I have with enums? or this may be an issue of kotlin.
12 replies
Hello everyone,
This April 27th, at 2:30 PM PT we are hosting a new gRPC community meetup. Please add this meeting to your calendar http://bit.ly/gRPCAprilMeetup and join us to hear working groups updates and ask any technical questions you may have. We will also show and discuss a demo, "ML Microservices with gRPC and Python", led by Andrej Baranovskij, Founder/Senior Developer at Katana ML.
Looking forward to seeing most of you this Thursday at our community meetup!
1 reply
Hello. I previously asked a question, what the difference between gRPC default executor and gRPC resolver executor is. But I didn't get any answers, So I ask it again.
In executor.cc#L401, when InitAll() calls, a pair of executor threads is always generated together.
executors[static_cast<size_t>(ExecutorType::DEFAULT)]->Init();
executors[static_cast<size_t>(ExecutorType::RESOLVER)]->Init();But I couldn't find any explanations about ExecutorType.
And I couldn't find whether two executor threads run different work.
2 replies
4 replies
14 UNAVAILABLE: Trying to connect an http1.x server error. I've put this c++ code into a fresh project using the same version of GRPC (1.12.0) and it works as expected. Comparing traces between the 2, the legacy project does not output any traces from ssl_transport_security but the fresh one does. This leads me to suspect that the SSL bits of the code is misbehaving in some way. Any help appreciated.
There was an error in the meetup link. The correct link is https://bit.ly/gRPCApril2021Meetup See you April 27th, at 2:30 PM PT For a discussion and a demo, "ML Microservices with gRPC and Python", led by Andrej Baranovskij, Founder/Senior Developer at Katana ML.
AuthMetadataProcessor (plain grpc-cpp) and hit the following issue. In the Process handler, there is an AuthContext that is passed. I'm noticing that in some cases, the AuthContext is reused between different calls on what I think are different channels. This means that the properties added with AddProperty are still present when handling what I think is a new call. I would have expected to get a "fresh" AuthContext for each channel (and my hypothesis here is: one new connection = one new channel). Am I misunderstanding something? Thanks.
4 replies
2 replies
1 reply
1 reply
Hi folks, I'm getting an ImportError with grpcio-1.37.0 for arm32 after installing it via PyPI (using grpcio-1.37.0-cp37-cp37m-linux_armv7l.whl).
import grpc
File "/usr/lib/python3.7/site-packages/grpc/__init__.py", line 23, in <module>
from grpc._cython import cygrpc as _cygrpc
ImportError: cannot import name 'cygrpc' from 'grpc._cython' (/usr/lib/python3.7/site-packages/grpc/_cython/__init__.py)I've double checked that grpcio-1.36.0 works fine on the same setup. Any clues about what may be going on? Not sure if this is an already known issue, I'm happy to open one on GitHub otherwise.
4 replies
${org}.some.thing eg. google.some.thing and uber.some.thing - is this established somewhere? I looked at the uber style guide (https://github.com/uber/prototool/blob/dev/style/README.md) and and google naming guide (https://cloud.google.com/apis/design/naming_convention) but neither of them explicitly make this suggestion
1 reply
Does anyone have any experience with taking a gRPC service implementation and putting it into a DLL, and some of the problems that might arise? I have successfully loaded a DLL that exposes a factory function to instantiate a Service Implementation (grpc::Service, for a PingPong proto w/ Unary request), and then I call builder.RegisterService(&service) successfully, but then I get some segfault on the server when I make the request and end up with an error like this:
Exception thrown: read access violation.
**grpc_core::g_static_metadata_slice_ctx** was nullptr.
Unhandled exception thrown: read access violation.
**grpc_core::g_static_metadata_slice_ctx** was nullptr.And it seems to just go on and on forever at least in debug mode w/ break points. Just curious if this is something others have encountered, I can suspect numerous potential problems:
1) some kind of issue with the DLL build of generated .protos interacting with the separate grpc-server.exe
2) DLL doesn't know how to properly access grpc::Server object's (PingPongServiceImplementation class) methods because DLL usage has scrambled them
Hi guys!, I'm @jmferrete, very nice to join this community.
I'm having some trouble with my GRPCLB round_robin configuration. I've defined my insecure_channel like this:
service_srv_dns_host_entry = "_grpclb._tcp.my.example.srv.entry.com"
opts = [("grpc.lb_policy_name", "round_robin"),
("grpc.dns_enable_srv_queries", True),
("grpc.max_connection_age_ms", 10000)]
channel = grpc.insecure_channel(f"dns:///{service_srv_dns_host_entry}", opts)And my SRV DNS entry resolves to three different backend DNS type A entries each of them with a different TCP port. Something like this:
0 0 50003 1.example.a.entry.com.
0 0 20003 2.example.a.entry.com.
0 0 30003 3.example.a.entry.com.If I use the insecure_channel, I can see this in the logs (extracted with GRPC_VERBOSITY=debug GRPC_TRACE=all):
I0428 10:54:44.048308771 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: input[0]=192.168.19.131:443
I0428 10:54:44.048386289 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: output[0]=192.168.19.131:443
I0428 10:54:44.048410145 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: input[0]=192.168.19.131:50003
I0428 10:54:44.048430944 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: input[1]=192.168.19.133:30003
I0428 10:54:44.048451026 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: input[2]=192.168.19.132:20003
I0428 10:54:44.048536903 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: output[0]=192.168.19.131:50003
I0428 10:54:44.048557022 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: output[1]=192.168.19.133:30003
I0428 10:54:44.048580054 22405 grpc_ares_wrapper.cc:559] (c-ares resolver) request:0x7f8980005a30 c-ares address sorting: output[2]=192.168.19.132:20003
I0428 10:54:44.049976937 22405 subchannel.cc:1012] Connect failed: {"created":"@1619600084.049891639","description":"Failed to connect to remote host: Connection refused","errno":111,"file":"src/core/lib/iomgr/tcp_client_posix.cc","file_line":201,"os_error":"Connection refused","syscall":"connect","target_address":"ipv4:192.168.19.131:443"}As we can see, it is trying to connect to the default TCP port (443), but we want to avoid it and request directly to the resolved TCP ports (50003, 20003, 30003). Am I missing or misunderstanding something? Thanks in advance.
5 replies
ClientInterceptor around an entire server-streaming RPC call? The server_streamer interceptor hook only wraps the initial establishment of the streaming connection, but I want to wrap the entire lifetime of the stream until it is closed.
grpc-node and want to know if it's a good idea to have a async/await style load balancer and picker, see grpc/grpc-node#1771
4 replies
Hi all, I'm trying to debug why we're getting client-side errors such as this with grpc-node.
Error: 14 UNAVAILABLE: GOAWAY received
at Object.exports.createStatusError (/code/node_modules/grpc/src/common.js:91:15)
at Object.onReceiveStatus (/code/node_modules/grpc/src/client_interceptors.js:1209:28)
at InterceptingListener._callNext (/code/node_modules/grpc/src/client_interceptors.js:568:42)
at InterceptingListener.onReceiveStatus (/code/node_modules/grpc/src/client_interceptors.js:618:8)
at callback (/code/node_modules/grpc/src/client_interceptors.js:847:24)
at callbackTrampoline (internal/async_hooks.js:126:14)Our server is setup as a headless Kubernetes service with horizontal auto scaling and the following option 'grpc.max_connection_age_ms': 30000 (so that the client will discover new replicas after connections expire).
Our client is setup with the following options.
'grpc.dns_min_time_between_resolutions_ms': 0,
'grpc.service_config': JSON.stringify({
loadBalancingConfig: [{
round_robin: {}
}]
})I've seen this issue also mentioned here.
https://github.com/grpc/grpc/issues/12295#issuecomment-650439430
I'm a bit confused why GOAWAY is treated as a error by the client, when it should signify a graceful connection termination, and I would expect transparent retries to take care of the resending of the request on a different connection?
https://github.com/grpc/proposal/blob/master/A9-server-side-conn-mgt.md
22 replies
v1.30.x, but the grpc-node native add-on version is pinned to v1.24.x)?grpc/grpc#23818