>Would injecting tcp rst really be necessary? In my test, doing "ip link del" 
>on the host side of the interface ALWAYS succeed, no matter
>what the state the guest container's interface is.
>
>Serge, do you have the particular commit ids for "lxc.network.script.down" 
>support? Backporting that would probably be the best step for
>me to try.

Dear Fajar, Dear Serge,

With my setup i found that on a test machine tcpdump detects a RST packet from 
the container (for an open, idle ssh connection to the containers sshd) only if 
the network interface of the container is *not* brought down at shutdown. 
Obviously, in that moment the ssh-client exists immediately with a "Connection 
to <container> closed by remote host". And I did not observe any "undeleted" 
veth's

If the interface is closed as usual, nothing happens at the test machine. Here 
the veth stay alive on the lxc-host. I'm able to remove the interface using 'ip 
link del dev <veth>'. This will allow to startup the container again using the 
same veth name (; i name it fixed by the container name). But in spite of this 
action, the ssh connection stay alive.

At the moment I don’t' have an idea where the "friendly" RST comes from. But it 
will terminate the tcp connections and therefore, the veth vanish at once.

Guido

------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from 
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60134791&iu=/4140/ostg.clktrk
_______________________________________________
Lxc-users mailing list
Lxc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/lxc-users

Reply via email to