Intermittent "context deadline exceeded" when no deadline set about go-libp2p HOT 7 CLOSED

The timeouts appear to be quite short in the occasions I observed -- maybe a couple seconds?

A little more detail about the specifics in what I have observed:

• All our connections use the context we created for the swarm, which is a background context with a cancellation, which we use when we want to shutdown networking. There is no timeout attached.
• I have noticed that when I hit the problem, closing the Host and creating a new instance also fails to connect. But if I let it retry (after 5 min -- it's a heartbeat for directory registrations in this particular case), it successfully connects. So the condition does not appear to be terminal.

from go-libp2p.

@parkan roughly how long are the timeouts? Having an approximate number would help narrow this down significantly

from go-libp2p.

Another data point, this time with an actual timeout measurement.
I am seeing repeated connection problems with dial attempt failed: context deadline exceeded, so I did a time measurement and it comes to around 10sec:

$ time mcclient id QmeiY2eHMwK92Zt6X4kUUC3MsjMmVb2VnGZ17DhnhRPCEQ
Not Found
Error: dial attempt failed: context deadline exceeded


real    0m10.498s
user    0m0.320s
sys 0m0.018s

All this is happening with the nodes involved (icmp) pinging happily and no unusual network conditions.

from go-libp2p.

Also verified that I can connect to the node with telnet and get the go-multistream header back.

from go-libp2p.

I also restarted the node I am trying to connect, and I still can't connect.

from go-libp2p.

Digging further, seems there is a legitimate networking problem with the node in this case (ssh yes, but no pings or telnet to the multistream port from my network location).

from go-libp2p.

I think we can ascribe my problems today at legitimate networking issues.

from go-libp2p.