Does Dgraph java client support gRPC retry?

Issues Dgraph Clients
,
1

Let’s say I setup the client like this (there are separate instances of Dgraph running on ports 9080 and 9083)

ManagedChannel channel1 = ManagedChannelBuilder
    .forAddress("localhost", 9080)
    .defaultLoadBalancingPolicy("round_robin")
    .defaultServiceConfig(serviceConfig)
    .enableRetry()
    .maxRetryAttempts(10)
    .usePlaintext().build();
DgraphStub stub1 = DgraphGrpc.newStub(channel1);

ManagedChannel channel2 = ManagedChannelBuilder
    .forAddress("localhost", 9083)
    .defaultLoadBalancingPolicy("round_robin")
    .defaultServiceConfig(serviceConfig)
    .enableRetry()
    .maxRetryAttempts(10)
    .usePlaintext().build();
    .usePlaintext().build();
DgraphStub stub2 = DgraphGrpc.newStub(channel2);

DgraphClient dgraphClient = new DgraphClient(stub1, stub2, stub3);

and my service config is:

{
  "methodConfig": [
    {
      "name": [
        {
          "service": "api.Dgraph"
        }
      ],

      "retryPolicy": {
        "maxAttempts": 5,
        "initialBackoff": "0.5s",
        "maxBackoff": "30s",
        "backoffMultiplier": 2,
        "retryableStatusCodes": [
          "UNAVAILABLE",
          "DEADLINE_EXCEEDED"
        ]
      }
    }
  ]
}

When both instances of Dgraph are up, I get a valid response to my queries. But, if I take one down, I just get a Connection refused exception (and UNAVAILABLE gRPC exception) when the request goes to that channel.
Note that, if the request goes to the channel still up, I do get a valid response.
With retry enabled, I would have expected the request to be sent to this channel one the first one fails.

For reference:

2

It seems that you need to implement the retry method yourself and switch to another channel when a connection error is caught.
And disable retry in the channel.

3

Yeah, that’s what I have concluded as well.
Doesn’t seem to support the gRPC retries.