https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing en a blog about Debian and self-hosting Fri, 06 Jul 2012 15:01:00 +0000 PluXml https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1341586886-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1341586886-1 @Marius Gedminas : I think this is new, but it is now possible to use port forwarding options with new clients that use an existing master connection, with the effect of adding these port forwards to the master connection. Fri, 06 Jul 2012 15:01:00 +0000 Tanguy https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1330105995-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1330105995-1 I've discovered another con not mentioned here that made me turn ControlMaster off: it disables the use of the ~C escape to add new port forwardings and whatnot. Fri, 24 Feb 2012 17:53:00 +0000 Marius Gedminas https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329727176-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329727176-1 @Tanguy: That's great! Finally! I guess I should've read it all the way through before I opened my mouth ;^) Thanks. Mon, 20 Feb 2012 08:39:00 +0000 rjc https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329608265-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329608265-1 Thank you for mentioning ControlPersist -- without it ControlMaster is just unusable (killing all the sessions when your randomly close the one xterm that just happens to contain the first one? Ouch). @friday: Ubuntu ships with a kernel security module (yama) that doesn't allow random processes ptrace each other (tweakable by a sysctl setting, if you have legitimate debugging needs). Fedora recently announced plans to reinvent that using a SELinux policy. Sat, 18 Feb 2012 23:37:00 +0000 Marius Gedminas https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329493117-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329493117-1 @Timo Juhani Lindfors : Excellent, I did not notice that option! Fri, 17 Feb 2012 15:38:00 +0000 Tanguy https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329493088-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329493088-1 @rjc : Well, this is exactly why I introduced the option “ControlPersist”, which forks and backgrounds the connection sharing thread. This way, the first client can terminate without bothering the other ones. Fri, 17 Feb 2012 15:38:00 +0000 Tanguy https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329475775-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329475775-1 anonymous, +1 I used to use this feature myself and the main disadvantage is the first connection being locked until the last one exits. Scenario: I logged on to a machine to do a quick maintenance task, then connected with another (this time under tmux) to do one lasting a bit longer. Not wanting to interrupt the latter I had to leave my screen locked as the first one wasn't run under tmux or screen. If only there was a way of the second (or third, fourth...) connection to take over as the main one, or better yet, for it to automatically "jump" until the last connection. Otherwise it's only useful as a detached job or interactive one locking one of your terminals. Fri, 17 Feb 2012 10:49:00 +0000 rjc https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329444109-1 https://tanguy.ortolo.eu/blog/article42/ssh-connection-sharing/#c1329444109-1 Let's extend the disadvantages a bit: * If you scp a file, it reuses your interactive connection and now sends your file using IPQos lowdelay clogging your bandwidth. Say good bye to interactive typing. * You just started scp and then log in interactively? Now your scp never finishes. This not only "works" with scp, but also with git, rsync and other tools. So maybe turning on this feature isn't that a good idea. Fri, 17 Feb 2012 02:01:00 +0000 anonymous