Process does not terminate after calling stop()

[Flamenco]

Issue

I am trying to shutdown my node by calling await helia.stop(), but there are still active timers and listening ports that prevent the process from terminating.

Workaround

call process.exit(0)

Another strategy that worked was to call gc() before stop(). Perhaps stop() should call gc() internally...

await helia.gc()
await helia.stop()

More Information

For your debugging pleasure I am including a dump of open items

open handles:
- File descriptors: (note: stdio always exists)
  - fd 2 (tty) (stdio)
  - fd 1 (tty) (stdio)
  - fd 0 (tty)
- Sockets:
  - 2600:4040:5b06:f300:7c33:3c4f:8cc3:26cf:52501 -> 2602:fea2:2::1:443
- Servers:
  - <unknown address>
    - Listeners:
      - connection: bound onSocket @ file:///Users/steven/dev/projects/ct/ct-labs/projects/helia-test/node_modules/@libp2p/tcp/src/listener.ts:88
  - <unknown address>
    - Listeners:
      - connection: bound onSocket @ file:///Users/steven/dev/projects/ct/ct-labs/projects/helia-test/node_modules/@libp2p/tcp/src/listener.ts:88
- Timers:
  - (5000 ~ 5 s) (anonymous) @ file:///Users/steven/dev/projects/ct/ct-labs/projects/helia-test/node_modules/libp2p/src/connection-manager/auto-dial.ts:107
  - (10000 ~ 10 s) (anonymous) @ node:internal/deps/undici/undici:8364
  - (100 ~ 100 ms) runHeartbeat @ file:///Users/steven/dev/projects/ct/ct-labs/projects/helia-test/node_modules/@chainsafe/libp2p-gossipsub/src/index.ts:598
Unable to determine callsite for "Function". Did you require `wtfnode` at the top of your entry point?
  - (2000 ~ 2 s) (anonymous) @ unknown:0
  - (1000 ~ 1000 ms) later @ file:///Users/steven/dev/projects/ct/ct-labs/projects/helia-test/node_modules/libp2p/src/address-manager/utils.ts:11
  - (84600000 ~ 23 hr) (anonymous) @ file:///Users/steven/dev/projects/ct/ct-labs/projects/helia-test/node_modules/p-timeout/index.js:72

[tuler]

Im also experiencing this, and calling gc didn’t help

[achingbrain]

Calling gc is almost certainly a red herring here, all that's going to do is delete any un-pinned blocks from your blockstore, it's not going to affect open TCP sockets.

The reported issue sounds similar to ChainSafe/lodestar#6053 which in turn seemed to be related to libp2p/js-libp2p#2058

That went out in @libp2p/tcp@8.0.8 - do you still see the issue if you use @libp2p/tcp@8.0.7 - if not we should probably revert that change.

[Flamenco]

Not planning on fixing a resource leak? Hmmmm.

[nflaig]

The reported issue sounds similar to ChainSafe/lodestar#6053 which in turn seemed to be related to libp2p/js-libp2p#2058

This is a different issue, see ChainSafe/lodestar#5642. Since we are running libp2p in a worker_thread now I can't tell if this is still there but seems related to ChainSafe/lodestar#5775 which is even more problematic.

[SgtPooki]

Not planning on fixing a resource leak? Hmmmm.

Normally this issue would be closed as stale because lack of author response. if you have response or remaining concerns, we'd be happy to look into it.