guacamole-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dogbert <>
Subject Re: Guacamole 0.9.12 WAN Performance (Ubuntu 14, XDRP)
Date Tue, 06 Jun 2017 18:42:13 GMT
Hi Mike,
Thank you so much.

>>> ...
>>>         GuacamoleSocket socket = new ConfiguredGuacamoleSocket(new
InetGuacamoleSocket("localhost", 4822), guacamoleConfig, info)
>>>         // Return a new tunnel which uses the connected socket
>>>         return new SimpleGuacamoleTunnel(socket)

Have you verified that this function is being called, and that things
aren't being routed to the HTTP tunnel straight off?

I have debugged through this and it does get called, how would I be able to
check the difference? In Firefox I found an extension that showed
web-socket traffic and that works, although on chrome is where I see a
bunch of get requests for images, for the same setup, but it seems to run
smoothly. Although it shouldn't make a difference, because I'm on centos 6,
I'm stuck at Chrome 49.

I will look into getting libwebp, and libjpeg-turbo installed, then look at
getting webp detection setup.

What you said about the CPU usage make sense to me, I was thinking along
those lines, but you put it into better words. With our setup it's typical
to see 400 running session, although I have no idea how many are actually
being used. In a production scenario, we might want to have multiple
webapps for socket connections, and multiple guacd. I know that my coworker
who worked on the novnc solution, currently has novnc, and websockify on
each node, being delegated from Apache.


On Tue, Jun 6, 2017 at 1:54 PM, Mike Jumper <>

> On Tue, Jun 6, 2017 at 10:33 AM, Dogbert <> wrote:
> > Thanks, actually for the error, it was just the session died, restarted
> it
> > and Guac is back. Still looking at the CPU spikes, upgrading didn't
> change
> > much. looking closer at the spikes, they usually take 80%, although that
> > would be out of 800%, seeing as I have four cores.with two threads per
> core.
> > So I'm not really sure if there is a CPU issue, or IT just over reacted.
> As long as the session is responsive, there is no issue. Instantaneous
> CPU usage is a poor indicator of performance for Guacamole, and you
> can expect that CPU will spike occasionally when complex changes are
> being made to the screen, or even consistently if those changes are
> happening continuously. The kernel does a rather good job of balancing
> load. Such spikes will only have practical implications if absolutely
> all users are constantly experiencing them and at the same time. In
> practice, humans are nondeterministic enough in their behavior that
> things tend not to intersect that badly unless there are quite a few
> users connected.
> Better indicators would be:
> 1) The load average over various intervals
> 2) A user's subjective experience on a different Guacamole connection
> on the same server at the same time
> - Mike

View raw message