This adds a mechanism for timing out a connection if the ssh handshake never completes.
When managing large fleets, we not-infrequently encounter cases where a host is alive enough to accept an SSH connection but not enough to accept commands. (I've seen this with failing disks, for example -- my hypothesis is enough of the ssh daemon is cached in memory, but the shell or other libraries can't be read.)
(Internal p4 CLN 5034409; patch approved for release via TT/0013249552)
Merge Amazon change 5034409 to GitHub:
Add a mechanism for timing out a connection if the ssh banner never completes.
@dacut sorry for the late reply! Do you think you can add som tests to this?
Expose handshake timeout in the transport API.
This is a reimplementation of #62.
Rolling into #491
Rework changelog entries re #491 a bit
Closes #491, closes #62, closes #439