Is it possible to use gRPC Client-side load balancing with a custom dns server?

[kharabasz]

Is it possible to use gRPC Client-side load balancing with a custom dns server, like the one bundled with consul? Unless I missed something in the code and docs, it looks like the current implementation goes to the system's dns server configuration when performing lookups.

[JamesNK]

The address with DNS is configurable. Can you create a GrpcChannel with DNS + consul's DNS address (I'm not familiar with DNS + consul)

[kharabasz]

What would be the dns uri that would correspond with the following lookup?

dig @consul-dns-server-address -p 8600 service-to-call.service.consul SRV

[kharabasz]

Reading through more of the dotnet documentation, I think the answer is no.

The Dns.* methods seem to rely on the internal Win32 APIs which in turn go through the DNS servers associated with the network connection. In order to get this to work, I would have to change the DNS servers associated with the network adapter's address - which I don't want to do.

I will try writing a custom DnsResolver against a package like DnsClient.NET

[kharabasz]

Using DnsClient.NET I put together a simple ConsulDnsResolver (code below).

This worked in our development environment until instances re-allocated, which did not trigger a refresh of the resolver. The error getting logged is:

Grpc.Core.RpcException: Status(
    StatusCode=\"Unavailable\", 
    Detail=\"Error connecting to subchannel.\", 
    DebugException=\"System.Net.Sockets.SocketException: Connection refused\"
)
       System.Net.Sockets.SocketException (111): Connection refused
       at System.Net.Sockets.Socket.AwaitableSocketAsyncEventArgs.ThrowException(SocketError error, CancellationToken cancellationToken)
       at System.Net.Sockets.Socket.AwaitableSocketAsyncEventArgs.System.Threading.Tasks.Sources.IValueTaskSource.GetResult(Int16 token)
       at System.Net.Sockets.Socket.\u003cConnectAsync\u003eg__WaitForConnectWithCancellation|277_0(AwaitableSocketAsyncEventArgs saea, ValueTask connectTask, CancellationToken cancellationToken)
       at Grpc.Net.Client.Balancer.Internal.SocketConnectivitySubchannelTransport.TryConnectAsync(ConnectContext context)\n   --- End of inner exception stack trace ---\n   at Grpc.Net.Client.Balancer.Internal.ConnectionManager.PickAsync(PickContext context, Boolean waitForReady, CancellationToken cancellationToken)
       at Grpc.Net.Client.Balancer.Internal.BalancerHttpHandler.SendAsync(HttpRequestMessage request, CancellationToken cancellationToken)
       at Grpc.Net.Client.Internal.GrpcCall`2.RunCall(HttpRequestMessage request, Nullable`1 timeout)
       at Grpc.Net.Client.Internal.Retry.RetryCallBase`2.GetResponseCoreAsync()
       at (domain code)

@JamesNK Off the top of your head - do you see anything glaringly wrong or missing from this code?

private static IServiceCollection RegisterGrpcClient(this IServiceCollection services, IConfiguration configuration)
{
    services.AddTransient<ILookupClient, LookupClient>(sp =>
    {
        var consulDnsServerIPAddress = IPAddress.Parse("127.0.0.1");
        return new LookupClient(consulDnsServerIPAddress, 8600);
    });

    services.AddSingleton<ResolverFactory>(sp =>
        new ConsulDnsResolverFactory(sp.GetRequiredService<ILookupClient>()));

    services
        .AddGrpcClient<ServiceToCallClient>(options =>
        {
            options.Address = new Uri("consul://service-to-call.service.consul");
            options.ChannelOptionsActions.Add(channelOptions =>
            {
                channelOptions.Credentials = ChannelCredentials.SecureSsl;
                channelOptions.ServiceConfig = new ServiceConfig
                {
                    MethodConfigs =
                    {
                        new MethodConfig
                        {
                            Names = { MethodName.Default },
                            RetryPolicy = new RetryPolicy
                            {
                                MaxAttempts = 3,
                                InitialBackoff = TimeSpan.FromSeconds(1),
                                MaxBackoff = TimeSpan.FromSeconds(5),
                                BackoffMultiplier = 1.5,
                                RetryableStatusCodes =
                                {
                                    StatusCode.Unavailable,
                                    StatusCode.Internal,
                                    StatusCode.Unknown,
                                    StatusCode.Cancelled,
                                    StatusCode.Aborted
                                }
                            }
                        }
                    },

                };
            });
        });
    
    return services;
}

internal class ConsulDnsResolver : PollingResolver
{
    private readonly ILookupClient _lookupClient;
    private readonly Uri _address;

    public ConsulDnsResolver(Uri address, ILoggerFactory loggerFactory, ILookupClient lookupClient) 
        : base(loggerFactory)
    {
        _address = address;
        _lookupClient = lookupClient;
    }

    protected override async Task ResolveAsync(CancellationToken cancellationToken)
    {
        try
        {
            var result = await _lookupClient.QueryAsync(
                _address.Host,
                QueryType.SRV,
                cancellationToken: cancellationToken);

            var balancerAddresses = new List<BalancerAddress>();
            foreach (var dnsResourceRecord in result.Answers)
            {
                if (dnsResourceRecord is not SrvRecord srvRecord) { continue; }

                foreach (var additionalResourceRecord in result.Additionals)
                {
                    if (additionalResourceRecord is not ARecord aRecord) continue;

                    if (aRecord.DomainName.Equals(srvRecord.Target))
                    {
                        balancerAddresses.Add(new BalancerAddress(aRecord.Address.ToString(), srvRecord.Port));
                    }
                }
            }

            if (!balancerAddresses.Any())
            {
                Listener(ResolverResult.ForFailure(new Status(StatusCode.NotFound, "No Consul DNS records found")));
            }
            else
            {
                Listener(ResolverResult.ForResult(balancerAddresses));
            }
        }
        catch (Exception e)
        {
            Listener(ResolverResult.ForFailure(new Status(StatusCode.Unknown, e.Message, e)));
        }
    }
}

internal class ConsulDnsResolverFactory : ResolverFactory
{
    private readonly ILookupClient _lookupClient;

    public ConsulDnsResolverFactory(ILookupClient lookupClient)
    {
        _lookupClient = lookupClient;
    }

    public override Resolver Create(ResolverOptions options)
    {
        return new ConsulDnsResolver(options.Address, options.LoggerFactory, _lookupClient);
    }

    public override string Name => "consul";
}

[JamesNK]

It looks ok to me.