Tiny binary for gRPC services for HEALTHCHECK docker command which calls the gRPC recommended health check service

Features

• Self contained binary
  • will work in FROM scratch images
  • same binary will work in most ^all? linux x86_64 based images
• Small - 2.3mb statically linked binary
• Quick installation using docker multi-build pattern

Prerequisites

Your gRPC server must have a service which implements the gRPC health checking protocol

<details> <summary>View .proto file:</summary> <!-- embedme proto/health_check.proto#L3-L24 -->

syntax = "proto3";

package grpc.health.v1;

message HealthCheckRequest {
    string service = 1;
}

message HealthCheckResponse {
    enum ServingStatus {
        UNKNOWN = 0;
        SERVING = 1;
        NOT_SERVING = 2;
    }
    ServingStatus status = 1;
}

service Health {
    rpc Check(HealthCheckRequest) returns (HealthCheckResponse);

    rpc Watch(HealthCheckRequest) returns (stream HealthCheckResponse);
}

</details>

Your Health service only needs to implement the unary Check rpc, as nightingale does not use the streaming endpoint. Feel free to throw 12: Unimplemented in the Watch endpoint.

Installation

In your Dockerfile add the following lines (this is using the multi stage build strategy)

COPY --from=xiphiaz/nightingale:latest /nightingale /nightingale

HEALTHCHECK CMD /nightingale

Nightingale Options

Note that options like --interval and --timeout should be passed to HEALTCHECK, NOT nightingale

<!-- embedme doc/help-output.txt#L5-L15 -->

USAGE:
    nightingale [OPTIONS] --host <host> --port <port>

FLAGS:
        --help       Prints help information
    -V, --version    Prints version information

OPTIONS:
    -h, --host <host>          Set host to use [default: [::1]]
    -p, --port <port>          Set port to use [default: 50051]
    -s, --service <service>    Which gRPC service request healthcheck for [default: ]

Comprehensive install:

COPY --from=xiphiaz/nightingale:latest /nightingale /nightingale

HEALTHCHECK --interval=5s --timeout=3s --start-period=10s --retries=5 CMD /nightingale --host 0.0.0.0 --port 50051 --service example.ExampleService

Usage

When your docker container starts, nightingale will be run every --interval (default 30s) which will in turn call the Check rpc endpoint of your Health service.

• If this returns SERVING, nightingale will exit 0 and the container is considered "healthy".
• If this returns NOT_SERVING or UNKNOWN nightingale will exit 1 and the container is considered "unhealthy".
• If this doesn't return at all, after --timeout (Default 30s), docker will cancel nightingale and consider the container "unhealthy".

You can verify the behavior and configuration is correct by running your container locally, and running the following command to view the docker-determined health status:

docker inspect --format='{{json .State.Health}}' $(docker ps -q  --filter ancestor=your-image-name) | jq

Note the interval configuration - you need to wait some time before seeing the docker log of health.

Future tasks

• Handle multiple service checks
• Clean up text output as docker inspect is very limiting in output format
• Unit tests over the input validation

Contributing

1. Install rust https://www.rust-lang.org/tools/install
2. Build & run

   
   rustup install beta
   
   rustup component add rustfmt --toolchain beta
   
   cargo +beta run --bin nightingale
   
   

3. Builder docker image

   docker build -t nightingale:latest .