ping unreachable node sometimes times out >1min

[alanshaw]

Sometimes this takes ~10s which is consistent with the code but it frequently takes >1 minute in CI:

[magik6k]

The real timeout is numPings*pingTimeout -
https://github.com/ipfs/go-ipfs/blob/31099e882469fab12a98f2069e38c847f0b1188c/core/commands/ping.go#L106

What is -n set to in those tests?

[alanshaw]

2 in that test https://github.com/ipfs/interface-ipfs-core/blob/0ca86fa081edfeb69be83d2a23fe2af71bfafde7/js/src/ping/ping.js#L54-L62

[alanshaw]

I thought that because the test is for an unreachable node it would actually drop in here https://github.com/ipfs/go-ipfs/blob/31099e882469fab12a98f2069e38c847f0b1188c/core/commands/ping.go#L90 and so the timeout would be 10s - which is consistent with the what we're seeing some of the time...

[Stebalien]

Triage: @alanshaw are you still seeing this?

and so the timeout would be 10s - which is consistent with the what we're seeing some of the time...

That's the timeout for connecting to the peer. There's another timeout below that's kPingTimeout * numPings. So this should total to kPingTimeout * (numPings+1) which should be 30s in your case.

So yeah, something is broken.

[alanshaw]

I've not removed the additional time we've allocated for the test to pass ;)

Looking at some recent CI builds I can't see it happening.

IMHO we should close this as multiple releases have happened since reporting, we've switched CI and now use the "test" profile in tests - any one of those factors (or something completely different) could have fixed this.

I can open a new issue if I notice it again.

That's the timeout for connecting to the peer. There's another timeout below...

Just FYI, I was not expecting it to get any further since this is for pinging an unreachable node. The Peer ID will not be in the peer store.

[Stebalien]

Yeah, I just want to make sure we don't have a weird deadlock.