test timeout in TestReconnect2 #1208
Labels
kind/bug
A bug in existing code (including security flaws)
Comments
marten-seemann
added
the
kind/bug
|
The problem here is the following: We have two hosts that are connected by a single TCP connection (the failure doesn't happen when using QUIC). When this connection is closed on the first host, in the vast majority of cases, the second host notices this quite quickly (the TCP connection closed, or rather, reset). Occasionally, the second host doesn't notice that the connection is closed though. Minimal reproducer: func TestConnectHang(t *testing.T) {
h1, err := bhost.NewHost(swarmt.GenSwarm(t, swarmt.OptDisableQUIC), nil)
require.NoError(t, err)
h2, err := bhost.NewHost(swarmt.GenSwarm(t, swarmt.OptDisableQUIC), nil)
require.NoError(t, err)
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
require.NoError(t, h1.Connect(ctx, peer.AddrInfo{ID: h2.ID(), Addrs: h2.Addrs()}))
require.Len(t, h1.Network().Conns(), 1)
require.Len(t, h2.Network().Conns(), 1)
h2.SetStreamHandler(protocol.TestingID, func(str network.Stream) {
io.ReadAll(str)
})
str, err := h1.NewStream(ctx, h2.ID(), protocol.TestingID)
require.NoError(t, err)
_, err = str.Write([]byte("foobar"))
require.NoError(t, err)
require.NoError(t, str.CloseWrite())
fmt.Println(h1.Network().Conns()[0])
require.NoError(t, h1.Network().Conns()[0].Close())
require.Eventually(t, func() bool { return len(h1.Network().Conns()) == 0 }, 10*time.Second, 10*time.Millisecond)
require.Eventually(t, func() bool { return len(h2.Network().Conns()) == 0 }, 10*time.Second, 10*time.Millisecond)
}output (that's the last The interesting thing is that both TCP RSTs are sent, so both connections should be closed. From a go routine dump it looks like the |
|
Interestingly, when I implement the same reproducer with two |
|
Update: This only fails on OSX. Here's a reproducer that only uses standard library packages: https://gist.github.com/marten-seemann/176b6bfc7115cf3984eb6756f35252cf. |
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The text was updated successfully, but these errors were encountered: