Cross-cluster search fails with skip_unavailable option

[prizov (Alexander Prizov)]

Hi everyone!

I have two remote clusters with skip_unavailable option. Cross-cluster search works fine.
But in case one of cluster is network unreachable I receive search timeout and don't get any response even from one cluster

Could you please explain why I faced with this problem?

Remote cluster configuration looks like this:

{
  "cluster1": {
    "seeds": [
      "cluster1-client-host1:9300",
      "cluster1-client-host2:9300"
    ],
    "http_addresses": [
      "10.200.100.2:9200",
      "10.200.100.3:9200",
      "10.200.100.4:9200"
    ],
    "connected": true,
    "num_nodes_connected": 3,
    "max_connections_per_cluster": 3,
    "initial_connect_timeout": "30s",
    "skip_unavailable": true
  },
  "cluster2": {
    "seeds": [
      "cluster2-client-host1:9300",
      "cluster2-client-host2:9300"
    ],
    "http_addresses": [
      "10.200.101.2:9200",
      "10.200.101.3:9200",
      "10.200.101.4:9200"
    ],
    "connected": true,
    "num_nodes_connected": 3,
    "max_connections_per_cluster": 3,
    "initial_connect_timeout": "30s",
    "skip_unavailable": true
  }
}

Thanks!

[javanna (Luca Cavanna)]

You seem to be setting skip_unavailable correctly, could you please the output of the _remote/info API? Also, do you configure CCS via config file or via API? And could you please post the exception that you get back?