Date: Fri, 23 Nov 2012 16:31:43 -0500 From: Lowell Gilbert <freebsd-questions-local@be-well.ilk.org> To: Paul Kraus <paul@kraus-haus.org> Cc: Carl Zwanzig <cpz@tuunq.com>, FreeBSD Mailing List <freebsd-questions@freebsd.org> Subject: Re: Odd X11 over SSH issue Message-ID: <44obioatvk.fsf@lowell-desk.lan> In-Reply-To: <CALhcXPDSZ8Qgj4tG1UQdO7c4tB3cSJbW%2Bj0DDbpJFct2VZXs4g@mail.gmail.com> (Paul Kraus's message of "Fri, 23 Nov 2012 12:43:58 -0500") References: <CALhcXPDSZ8Qgj4tG1UQdO7c4tB3cSJbW%2Bj0DDbpJFct2VZXs4g@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Paul Kraus <paul@kraus-haus.org> writes: > I am seeing very poor response time running the VitrualBox GUI via X11 > tunneled over SSH via the Internet. The issue _appears_ to be limited > to the VBox GUI as Firefox is reasonable. I am well aware of the > latency issues tunneling X11 over SSH across the Internet, but that is > what we are stuck with for the moment. The server is running FreeBSD 9 > and is patched as of about 4 weeks ago. > > Observations: > > 1. When I first SSH into the box I see a long delay after the SSH > tunnel is setup before being prompted for a password, and I do not > know if this delay is related to the VBox issue. Details below. Running the ssh server with more debugging will probably tell you what's happening in this area. > 2. When I fire up VirtualBox it takes _minutes_ before the window > opens and each action (drawing contents, mouse clicks) takes > additional _minutes_. Looking at the VirtualBox process with truss I > see many, many errors of the form: > > read(7,0x80193a02c,4096) ERR#35 'Resource temporarily unavailable' > > where fd 7 is a socket. This could be a red herring. Or not. But you can't tell without tracing down exactly what the socket is, and what is expected to be read from it. Probably not the first path worth exploring, although you may need to go there eventually. > I would chock it up to network slowness, but I > do not see the same behavior with Firefox, xload, or xclock. That's not a fair comparison, because tunneling a whole X server involves passing a lot more events than tunneling an application to run on your local server. This is particularly painful because the X protocols are highly serial. Is there any particular reason you don't let the X server run remotely and attach to it with something more latency-friendly, like vnc? I would expect that to work vastly better on any OS, just because you get X (specifically, its tendency to head-of-line blocking) out of its own way. Be well.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?44obioatvk.fsf>