Before you start writing a test, you could also look at the sshd 0.10.0 candidate release for which I started a vote yesterday. There are some important changes when closing which may have impacted what you've seen (and maybe even fixed it ?).
2014-02-11 0:48 GMT+01:00 Darren Shepherd <[email protected]>: > Sure I'll give that a shot. I haven't looked at the unit tests so it > might take me a bit to write a test. > > Darren > > > On Feb 10, 2014, at 1:49 PM, Guillaume Nodet <[email protected]> wrote: > > > > Do you think you can raise a JIRA and write a junit test so that I can > have > > a look ? > > > > > > 2014-02-10 18:13 GMT+01:00 Darren Shepherd <[email protected] > >: > > > >> I'm doing a lot of reverse forwarding and ran into a situation in which > I > >> thought I was leaking sockets. If you do a reverse forward and then > just > >> keep connecting and disconnecting to the port the number of open sockets > >> just keeps going up. > >> > >> What I figured out is that it's not really going up it just takes awhile > >> to close. If you connect to the port every 5 seconds it will level off > at > >> a couple hundred. The ssh server sends a channel eof but the socket > >> doesn't get released until some other thread comes along that seems to > >> clean up sessions. I don't fully understand the details of that. > >> > >> What I've done to work around this issue is just close the sockets on > >> channel eof. I'm not really sure if that is a generally safe change, > but > >> it is working for my use case. > >> > >> I'm not sure what a proper fix would be or even if this is consider a > bug, > >> but the end result of the behavior is that reverse forwarding is not > useful > >> when you expect to have connections happening regularly. You will > quickly > >> run out of FDs. > >> > >> Darren >
