Gateways stall after 5 minutes when using reframe/delegated routing

[ajnavarro]

Version

v0.17.0

Config

"Routing": {
    "Methods": {
      "find-peers": {
        "RouterName": "DHTs"
      },
      "find-providers": {
        "RouterName": "ParallelHelper"
      },
      "get-ipns": {
        "RouterName": "DHTs"
      },
      "provide": {
        "RouterName": "DHTs"
      },
      "put-ipns": {
        "RouterName": "DHTs"
      }
    },
    "Routers": {
      "CidContact": {
        "Parameters": {
          "Endpoint": "https://cid.contact/reframe"
        },
        "Type": "reframe"
      },
      "ParallelHelper": {
        "Parameters": {
          "Routers": [
            {
              "IgnoreErrors": true,
              "RouterName": "CidContact",
              "Timeout": "3s"
            },
            {
              "IgnoreErrors": false,
              "RouterName": "DHTs",
              "Timeout": "5m"
            }
          ]
        },
        "Type": "parallel"
      },
      "DHTs": {
        "Parameters": {
          "Routers": [
            {
              "IgnoreErrors": false,
              "RouterName": "WanDHT",
              "Timeout": "5m"
            },
            {
              "IgnoreErrors": false,
              "RouterName": "LanDHT",
              "Timeout": "1m"
            }
          ]
        },
        "Type": "parallel"
      },
      "WanDHT": {
        "Parameters": {
          "AcceleratedDHTClient": true,
          "Mode": "auto",
          "PublicIPNetwork": true
        },
        "Type": "dht"
      },
      "LanDHT": {
        "Parameters": {
          "AcceleratedDHTClient": false,
          "Mode": "auto",
          "PublicIPNetwork": false
        },
        "Type": "dht"
      }
    },

    "Type": "custom"
  },

Description

After ~5 minutes of having the following configuration, delegated routing stops sending find providers requests.

Might be a deadlock problem. A possible solution here: libp2p/go-libp2p-routing-helpers#64

[BigLep]

@ajnavarro @guseggert : I'm reopening because:

1. I don't see a regression test anywhere (looking at Return a closed channel instead of nil for getChannelOrErrorParallel libp2p/go-libp2p-routing-helpers#64 and chore: Update go-routing-helpers dependency #9447 ). How do we know this fixes the issue?
2. Should we get chore: Add logs to composable parallel libp2p/go-libp2p-routing-helpers#63 merged as weel?

[guseggert]

I added a test here: libp2p/go-libp2p-routing-helpers#66

We should get the logging change merged too but I don't think it's necessary to unblock the gateway deployment

[guseggert]

Fix has been merged & deployed, I verified that the deadlock is no longer occurring, and there's a regression test.