Hi, I have been trying to resolve some issues where connection drops. I have had limited success by doing this - https://metalbear.co/mirrord/docs/faq/common-issues#agent-connection-fails-or-drops-when-using-an-ephemeral-agent-with-a-service-mesh The first time it works, but the second time I get this.

{
  "timestamp": "2025-07-18T05:54:45.794723Z",
  "level": "ERROR",
  "fields": {
    "error": "io error: Address already in use (os error 98)"
  },
  "target": "mirrord_agent::entrypoint",
  "span": {
    "args": "Args { mode: Ephemeral { mesh: None }, communicate_port: 51234, communication_timeout: 360, network_interface: None, metrics: None, test_error: false, operator_tls_cert_pem: None, is_mesh: true, ipv6: false }",
    "name": "start_agent"
  },
  "spans": [
    {
      "args": "Args { mode: Ephemeral { mesh: None }, communicate_port: 51234, communication_timeout: 360, network_interface: None, metrics: None, test_error: false, operator_tls_cert_pem: None, is_mesh: true, ipv6: false }",
      "name": "start_agent"
    }
  ],
  "threadId": "ThreadId(1)"
}

I don't have

ttl

set, but I do have

communication_timeout: 360
      startup_timeout: 360

Will these keep the agent alive? Or is there a recommended way to make this work? Thanks

Hi Brad, It seems that you have a lingering agent on the target Can you make sure you don’t have any local mirrord processes running before trying to use mirrord?

The first ephemeral container (the working one) has logged this

│ {"timestamp":"2025-07-18T06:01:46.460798Z","level":"TRACE","fields":{"message":"start_agent -> All clients finished, exitin │
│ g main agent loop"},"target":"mirrord_agent::entrypoint","span":{"args":"Args { mode: Ephemeral { mesh: None }, communicate │
│ _port: 51234, communication_timeout: 360, network_interface: None, metrics: None, test_error: false, operator_tls_cert_pem: │
│  None, is_mesh: true, ipv6: false }","name":"start_agent"},"spans":[{"args":"Args { mode: Ephemeral { mesh: None }, communi │
│ cate_port: 51234, communication_timeout: 360, network_interface: None, metrics: None, test_error: false, operator_tls_cert_ │
│ pem: None, is_mesh: true, ipv6: false }","name":"start_agent"}],"threadId":"ThreadId(1)"}                                   │
│ {"timestamp":"2025-07-18T06:01:46.460804Z","level":"TRACE","fields":{"message":"start_agent -> Agent shutting down, droppin │
│ g cancellation token for background tasks"},"target":"mirrord_agent::entrypoint","span":{"args":"Args { mode: Ephemeral { m │
│ esh: None }, communicate_port: 51234, communication_timeout: 360, network_interface: None, metrics: None, test_error: false │
│ , operator_tls_cert_pem: None, is_mesh: true, ipv6: false }","name":"start_agent"},"spans":[{"args":"Args { mode: Ephemeral │
│  { mesh: None }, communicate_port: 51234, communication_timeout: 360, network_interface: None, metrics: None, test_error: f │
│ alse, operator_tls_cert_pem: None, is_mesh: true, ipv6: false }","name":"start_agent"}],"threadId":"ThreadId(1)"}

So I thought that it meant that there were no lingering processes

Makes sense

{
  "target": {
    "path": "{{ get_env(name='MD_DEPLOYMENT', default='deployment/playmaker-grpc/container/grpc') }}",
    "namespace": "{{ get_env(name='MD_NAMESPACE', default='') }}"
  },
  "feature": {
    "network": {
      "incoming": {
        "mode": "{{ get_env(name='MD_MODE', default='off') }}",
        "ignore_ports": [
          8081
        ],
        "port_mapping": [
          [
            50000,
            5000
          ]
        ]
      },
      "outgoing": true
    },
    "fs": {
      "mode": "read",
      "local": [
        ".+\\.js",
        ".+\\.mjs",
        ".+\\.ts",
        ".+\\.go",
        "dist/*"
      ]
    },
    "env": {
      "override": {
        "GRPC_PORT": "{{ get_env(name='MD_GRPC_PORT', default='50000') }}",
        "GRPC_REFLECTION": "true"
      }
    }
  },
  "operator": true,
  "skip_processes": "tsc",
  "kube_context": "{{ get_env(name='MD_CLUSTER', default='dev') }}",
  "internal_proxy": {
    "log_destination": "/tmp/mirrord.log",
    "log_level": "mirrord=info"
  }
}

Yes, have the agent to a static port as per https://metalbear.co/mirrord/docs/faq/common-issues#agent-connection-fails-or-drops-when-using-an-ephemeral-agent-with-a-service-mesh So we can exclude it from istio

traffic.sidecar.istio.io/excludeInboundPorts: 'xxxx
'

Oh you don’t need to. There’s a flag we added to auto resolve that, one sec

Thankyou! I'll give it a shot