grpcdebug is a command line interface focusing on simplifying the debugging process of gRPC applications. grpcdebug fetches the internal states of the gRPC library from the application via gRPC protocol and provide a human-friendly UX to browse them. Currently, it supports Channelz/Health Checking/CSDS (aka. admin services). In other words, it can fetch statistics about how many RPCs has being sent or failed on a given gRPC channel, it can inspect address resolution results, it can dump the in-effective xDS configuration that directs the routing of RPCs.
If you are looking for a tool to send gRPC requests and interact with a gRPC server, please checkout https://github.com/fullstorydev/grpcurl.
grpcdebug is an gRPC service admin CLI
Usage:
grpcdebug <target address> [flags] <command>
Available Commands:
channelz Display gRPC states in a human readable way.
health Check health status of the target service (default "").
help Help about any command
xds Fetch xDS related information.
Flags:
--credential_file string Sets the path of the credential file; used in [tls] mode
-h, --help help for grpcdebug
--security string Defines the type of credentials to use [tls, google-default, insecure] (default "insecure")
--server_name_override string Overrides the peer server name if non empty; used in [tls] mode
-t, --timestamp Print timestamp as RFC3339 instead of human readable strings
-v, --verbose Print verbose information for debugging
Use "grpcdebug <target address> [command] --help" for more information about a command.
- grpcdebug
The download links of the binaries can be found at
https://github.com/grpc-ecosystem/grpcdebug/releases. You can find the
precompiled artifacts for macOS
/Linux
/Windows
.
Minimum Golang Version 1.15. Official Golang install guide: https://golang.org/doc/install.
You can install the grpcdebug
tool using command:
# For Golang 1.16+
go install -v github.com/grpc-ecosystem/grpcdebug@latest
# For Golang 1.15
GO111MODULE=on go get -v github.com/grpc-ecosystem/grpcdebug
You can check your Golang version with:
go version
Don't forget to add Golang binaries to your PATH
:
export PATH=$PATH:$(go env GOPATH)/bin
If certain commands are confusing, please try to use -h
to get more context.
Suggestions and ideas are welcome, please post them to
https://github.com/grpc-ecosystem/grpcdebug/issues!
If you haven't got your gRPC application instrumented, feel free to try out the
mocking testserver
which implemented admin services.
cd internal/testing/testserver
go run main.go
# Serving Business Logic on :10001
# Serving Insecure Admin Services on :50051
# Serving Secure Admin Services on :50052
# ...
To connect to a gRPC endpoint without any credentials, we don't use any special
flags. If the local network can connect to the given gRPC endpoint, it should
just work. For example, if I have a gRPC application exposing admin services at
localhost:50051
:
grpcdebug localhost:50051 channelz channels
One way to establish a TLS connection with grpcdebug is by specifying the credentials via command line flags. For example:
grpcdebug localhost:50052 --security=tls --credential_file=./internal/testing/ca.pem --server_name_override="*.test.youtube.com" channelz channels
Alternatively, like OpenSSH clients, you can specify the security settings in a
grpcdebug_config.yaml
file. grpcdebug CLI will find matching connection config and
then use it to connect.
servers:
"pattern string":
real_address: string
security: insecure/tls
credential_file: string
server_name_override: string
Here is an example config file grpcdebug_config.yaml.
Each server config can have the following settings:
- Pattern: the string right after
Server
which dictates if this rule should apply; - RealAddress: if present, override the given target address, which allows giving nicknames/aliases to frequently used addresses;
- Security: allows
insecure
ortls
, expecting more in the future; - CredentialFile: path to the credential file;
- ServerNameOverride: override the hostname, which is useful for local reproductions to comply with the certificates' common name requirement.
grpcdebug searches the config file in the following order:
- Check if the environment variable
GRPCDEBUG_CONFIG
is set, if so, load from the given path; - Try to load the
grpcdebug_config.yaml
file in the current working directory; - Try to load the
grpcdebug_config.yaml
file in the user config directory (Linux:$HOME/.config
, macOS:$HOME/Library/Application Support
, Windows:%AppData%
, seeos.UserConfigDir()
).
For example, we can connect to our mock test server's secure admin port via:
GRPCDEBUG_CONFIG=internal/testing/grpcdebug_config.yaml grpcdebug localhost:50052 channelz channels
# Or
GRPCDEBUG_CONFIG=internal/testing/grpcdebug_config.yaml grpcdebug prod channelz channels
grpcdebug can be used to fetch the health checking status of a peer gRPC
application (see
health.proto).
gRPC's health checking works at the service-level, meaning services registered on
the same gRPC server may have different health statuses. The health status of
service ""
is used to represent the overall health status of the gRPC
application.
To simply fetch the overall health status:
grpcdebug localhost:50051 health
# <Overall>: SERVING
# or
# <Overall>: NOT_SERVING
Or fetch individual service's health status:
grpcdebug localhost:50051 health helloworld.Greeter
# <Overall>: SERVING
# helloworld.Greeter: SERVING
Channelz is a channel tracing library that allows applications to remotely query gRPC internal debug information. Also, Channelz has a web interface (see gdebug). grpcdebug is able to fetch information and present it in a more readable way.
Generally, you wil start with either the servers
or channels
command and
then work down to the details.
For all Channelz commands, you can add --json
to get the raw Channelz output.
grpcdebug localhost:50051 channelz servers --json
#[
# {
# "ref": {
# "server_id": 2,
# "name": "ServerImpl{logId=2, transportServer=NettyServer{logId=1, addresses=[0.0.0.0/0.0.0.0:50051]}}"
# },
# "data": {
# "calls_started": 3,
# "calls_succeeded": 2,
# "last_call_started_timestamp": {
# "seconds": 1680220688,
# "nanos": 444000000
# }
# },
# "listen_socket": [
# {
# "socket_id": 3,
# "name": "ListenSocket{logId=3, channel=[id: 0x05f9f16c, L:/0:0:0:0:0:0:0:0%0:50051]}"
# }
# ]
# }
#]
grpcdebug localhost:50051 channelz channels
# Channel ID Target State Calls(Started/Succeeded/Failed) Created Time
# 7 localhost:10001 READY 5136/4631/505 8 minutes ago
grpcdebug localhost:50051 channelz servers
# Server ID Listen Addresses Calls(Started/Succeeded/Failed) Last Call Started
# 1 [:::10001] 2852/2530/322 now
# 2 [:::50051] 29/28/0 now
# 3 [:::50052] 4/4/0 26 seconds ago
You can identify a channel via the Channel ID.
grpcdebug localhost:50051 channelz channel 7
# Channel ID: 7
# Target: localhost:10001
# State: READY
# Calls Started: 3976
# Calls Succeeded: 3520
# Calls Failed: 456
# Created Time: 6 minutes ago
# ---
# Subchannel ID Target State Calls(Started/Succeeded/Failed) CreatedTime
# 8 localhost:10001 READY 3976/3520/456 6 minutes ago
# ---
# Severity Time Child Ref Description
# CT_INFO 6 minutes ago Channel Created
# CT_INFO 6 minutes ago Resolver state updated: {Addresses:[{Addr:localhost:10001 ServerName: Attributes:<nil> Type:0 Metadata:<nil>}] ServiceConfig:<nil> Attributes:<nil>} (resolver returned new addresses)
# CT_INFO 6 minutes ago Channel switches to new LB policy "pick_first"
# CT_INFO 6 minutes ago subchannel(subchannel_id:8 ) Subchannel(id:8) created
# CT_INFO 6 minutes ago Channel Connectivity change to CONNECTING
# CT_INFO 6 minutes ago Channel Connectivity change to READY
grpcdebug localhost:50051 channelz subchannel 8
# Subchannel ID: 8
# Target: localhost:10001
# State: READY
# Calls Started: 4490
# Calls Succeeded: 3966
# Calls Failed: 524
# Created Time: 7 minutes ago
# ---
# Socket ID Local->Remote Streams(Started/Succeeded/Failed) Messages(Sent/Received)
# 9 ::1:47436->::1:10001 4490/4490/0 4490/3966
grpcdebug localhost:50051 channelz socket 9
# Socket ID: 9
# Address: ::1:47436->::1:10001
# Streams Started: 4807
# Streams Succeeded: 4807
# Streams Failed: 0
# Messages Sent: 4807
# Messages Received: 4243
# Keep Alives Sent: 0
# Last Local Stream Created: now
# Last Remote Stream Created: a long while ago
# Last Message Sent Created: now
# Last Message Received Created: now
# Local Flow Control Window: 65535
# Remote Flow Control Window: 65535
# ---
# Socket Options Name Value
# SO_LINGER [type.googleapis.com/grpc.channelz.v1.SocketOptionLinger]:{duration:{}}
# SO_RCVTIMEO [type.googleapis.com/grpc.channelz.v1.SocketOptionTimeout]:{duration:{}}
# SO_SNDTIMEO [type.googleapis.com/grpc.channelz.v1.SocketOptionTimeout]:{duration:{}}
# TCP_INFO [type.googleapis.com/grpc.channelz.v1.SocketOptionTcpInfo]:{tcpi_state:1 tcpi_options:7 tcpi_rto:204000 tcpi_ato:40000 tcpi_snd_mss:32768 tcpi_rcv_mss:1093 tcpi_last_data_sent:16 tcpi_last_data_recv:16 tcpi_last_ack_recv:16 tcpi_pmtu:65536 tcpi_rcv_ssthresh:65476 tcpi_rtt:192 tcpi_rttvar:153 tcpi_snd_ssthresh:2147483647 tcpi_snd_cwnd:10 tcpi_advmss:65464 tcpi_reordering:3}
# ---
# Security Model: TLS
# Standard Name: TLS_AES_128_GCM_SHA256
grpcdebug localhost:50051 channelz server 1
# Server Id: 1
# Listen Addresses: [:::10001]
# Calls Started: 5250
# Calls Succeeded: 4647
# Calls Failed: 603
# Last Call Started: now
# ---
# Socket ID Local->Remote Streams(Started/Succeeded/Failed) Messages(Sent/Received)
# 10 ::1:10001->::1:47436 5250/5250/0 4647/5250
In production, there may be thousands of clients/servers/sockets. It would be very noisy to print all of them at once, so Channelz supports pagination through start_id
and max_results
grpcdebug localhost:50051 channelz servers --start_id=0 --max_results=1
# Server ID Listen Addresses Calls(Started/Succeeded/Failed) Last Call Started
# 1 [:::10001] 2852/2530/322 now
grpcdebug localhost:50051 channelz servers --start_id=2 --max_results=2
# Server ID Listen Addresses Calls(Started/Succeeded/Failed) Last Call Started
# 2 [:::50051] 29/28/0 now
# 3 [:::50052] 4/4/0 26 seconds ago
It works similarly for printing channels via channelz channels
and printing server sockets via channelz server
.
xDS is a data plane configuration API commonly used in service mesh projects. It's created by Envoy, used by Istio, Traffic Director, and gRPC.
The xDS resources status might be REQUESTED
/DOES_NOT_EXIST
/ACKED
/NACKED
(see
config_dump.proto).
This view is intended for a quick scan if a configuration is propagated from the
service mesh control plane.
grpcdebug localhost:50051 xds status
# Name Status Version Type LastUpdated
# xds-test-server:1337 ACKED 1617141154495058478 type.googleapis.com/envoy.config.listener.v3.Listener 2 days ago
# URL_MAP/1040920224690_sergii-psm-test-url-map_0_xds-test-server:1337 ACKED 1617141154495058478 type.googleapis.com/envoy.config.route.v3.RouteConfiguration 2 days ago
# cloud-internal-istio:cloud_mp_1040920224690_6530603179561593229 ACKED 1617141154495058478 type.googleapis.com/envoy.config.cluster.v3.Cluster 2 days ago
# cloud-internal-istio:cloud_mp_1040920224690_6530603179561593229 ACKED 1 type.googleapis.com/envoy.config.endpoint.v3.ClusterLoadAssignment 2 days ago
grpcdebug localhost:50051 xds config
# {
# "config": [
# {
# "node": {
# "id": "projects/1040920224690/networks/default/nodes/5cc9170c-d5b4-4061-b431-c1d43e6ac0ab",
# "cluster": "cluster",
# "metadata": {
# "INSTANCE_IP": "192.168.120.31",
# "TRAFFICDIRECTOR_GCP_PROJECT_NUMBER": "1040920224690",
# "TRAFFICDIRECTOR_NETWORK_NAME": "default"
# },
# ...
For an example config dump, see csds_config_dump.json.
The dumped xDS config can be quite verbose, if I only interested in certain xDS type, grpcdebug can only print the selected section.
grpcdebug localhost:50051 xds config --type=eds
# {
# "dynamicEndpointConfigs": [
# {
# "versionInfo": "1",
# "endpointConfig": {
# "@type": "type.googleapis.com/envoy.config.endpoint.v3.ClusterLoadAssignment",
# "clusterName": "cloud-internal-istio:cloud_mp_1040920224690_6530603179561593229",
# "endpoints": [
# {
# "locality": {
# "subZone": "jf:us-central1-a_7062512536751318190_neg"
# },
# "lbEndpoints": [
# {
# "endpoint": {
# "address": {
# "socketAddress": {
# "address": "192.168.120.26",
# "portValue": 8080
# }
# }
# },
# "healthStatus": "HEALTHY"
# }
# ],
# "loadBalancingWeight": 100
# }
# ]
# },
# "lastUpdated": "2021-03-31T01:20:33.936Z",
# "clientStatus": "ACKED"
# }
# ]
# }
--- a/examples/src/main/java/io/grpc/examples/helloworld/HelloWorldServer.java
+++ b/examples/src/main/java/io/grpc/examples/helloworld/HelloWorldServer.java
@@ -18,6 +18,7 @@ package io.grpc.examples.helloworld;
import io.grpc.Server;
import io.grpc.ServerBuilder;
+import io.grpc.services.AdminInterface;
import io.grpc.stub.StreamObserver;
import java.io.IOException;
import java.util.concurrent.TimeUnit;
@@ -36,6 +37,7 @@ public class HelloWorldServer {
int port = 50051;
server = ServerBuilder.forPort(port)
.addService(new GreeterImpl())
+ .addServices(AdminInterface.getStandardServices())
.build()
.start();
logger.info("Server started, listening on " + port);
--- a/examples/helloworld/greeter_server/main.go
+++ b/examples/helloworld/greeter_server/main.go
@@ -27,6 +27,7 @@ import (
"net"
"google.golang.org/grpc"
+ "google.golang.org/grpc/admin"
pb "google.golang.org/grpc/examples/helloworld/helloworld"
)
@@ -51,6 +52,11 @@ func main() {
log.Fatalf("failed to listen: %v", err)
}
s := grpc.NewServer()
+ cleanup, err := admin.Register(s)
+ if err != nil {
+ log.Fatalf("failed to register admin: %v", err)
+ }
+ defer cleanup()
pb.RegisterGreeterServer(s, &server{})
if err := s.Serve(lis); err != nil {
log.Fatalf("failed to serve: %v", err)
--- a/examples/cpp/helloworld/greeter_server.cc
+++ b/examples/cpp/helloworld/greeter_server.cc
@@ -20,6 +20,7 @@
#include <memory>
#include <string>
+#include <grpcpp/ext/admin_services.h>
#include <grpcpp/ext/proto_server_reflection_plugin.h>
#include <grpcpp/grpcpp.h>
#include <grpcpp/health_check_service_interface.h>
@@ -60,6 +61,7 @@ void RunServer() {
// Register "service" as the instance through which we'll communicate with
// clients. In this case it corresponds to an *synchronous* service.
builder.RegisterService(&service);
+ grpc::AddAdminServices(&builder);
// Finally assemble the server.
std::unique_ptr<Server> server(builder.BuildAndStart());
std::cout << "Server listening on " << server_address << std::endl;