Forum Replies Created
I think this has been resolved after a proper re-install of our Qt app: the Qt libs shipped with the product were being found it turned out (there was no fallback), but there were other unresolved libraries that are now found.
And the double-click responsiveness was a separate issue–I had to change the threshold in the xfce-perchannel-xml file.
Thanks fisherman, I had seen that (but reviewed it again). I was thinking about a fixed workload rather, maybe even an internal (NoMachine) regression test suite, that could be run either from the web client (e.g. a button below ‘Take the statistics’ or via the nxwebplayer URL, simulating sets of requests at the browser end.
Eventually, I’d like to build my own regression suite for my applications, but in the mean time, I thought something like this might be useful to the broader community ? And even NoMachine support maybe, in looking at performance issues.
Good lead … the Qt libs are unresolved (ldd reports “not found”).
I’m pretty sure I didn’t install this app properly come to think, and I didn’t install Qt I’m sure, though it more or less works otherwise (maybe it has a fallback if Qt isn’t available ?). This aligns with us not seeing this problem outside of this NoMachine deployment.
Comparing to a properly installed version btw, ldd reports libQt* v4, which resolves to v4.6.2.
I’ll reinstall and retest.
Correction about the test client machine: Windows 7 Pro (SP1) + Firefox 52.5.3 (32-bit) with 8GB memory.
Its with our own Qt-based app–its a fairly mature product.
Good point, it occurred to me yesterday that a baseline would be useful too, I started playing with Xfce’s terminal app: no signs of the problem there but I might not have been using it for long enough for the problem to show up.
I’d move CPU down on the list of potential culprits, as the same basic operations were done by two/three different users at the customer’s office–with similar results (we hear).
The desktop environment & apps are the same for all users.
As for network latency, I’d taken our reasonable performance between London and Toronto as an lower limit to what I’d expect between our London DC and the customer’s London office. Bandwidth I hadn’t really considered though, I’d assumed it was enough, with YouTube and all … something to confirm though.
Since I posted originally, our customer has engaged their IT support to investigate further, as they tried a session from outside their office’s network and saw the (good) performance they were expecting. I haven’t received any feedback yet, but if they come back to us I’ll probably ask them for their workstation resources and internet bandwidth, maybe ask them to run some benchmarks.
Thanks for the ideas, some good things to think about. Its feeling less-and-less like a NoMachine issue.
Check the ClientConnectionMethods parameter in NX/etc/server.cfg (node.cfg). I think it should include “HTTP”.
– ValDecember 16, 2015 at 12:50 in reply to: Cloud Server v5 – embedding NoMachine, customer branding #9355
Thanks Britgirl, I’ll follow through with the Sales team when ready.
Thanks NiKO … feel free to let me know if you need any more information.
ValDecember 3, 2015 at 10:03 in reply to: Cloud Server v5, NoMachine ‘Connection to the server was lost’ after Xfce logout #9233
Thanks very much, haven. Looking forward to the fix, which we would need before we can proceed towards a production deployment.
Fwiw, I realized the xfwm4 process I had been running from ~/.xsession is no longer needed now that I’m running startxfce4 (as opposed to running some X11 app directly, which I had been doing before).
Also, I expect this goes without saying, but in addition to the re-login being made to work (from the same web browser session) after the Xfce session is closed, I think the at-exit message shown to the user, ‘The connection to the server was lost’ could be re-worded so the user doesn’t think there is a problem, e.g. ‘The connection to the server is closed.’ (without “Ooops”), or possibly even return the user to the prior NoMachine screen with the Existing/New Desktop choices, i.e. not requiring the user to press [OK].
– ValDecember 2, 2015 at 11:06 in reply to: Non-default install path & nxexec gives “Could not retrieve NX scripts directory” #9219
Thanks graywolf, I have uninstalled and reinstalled NoMachine (v5.0.53) and have not tweaked the paths this time, though I have installed at my chosen paths (/opt/nomachine/NX/, /etc/opt/nomachine/NX/, /var/opt/nomachine/NX/) with symlinks at the default paths (/usr/NX/, /etc/NX/, /var/NX/) this time.
Fwiw, if you’ve ever considered changing the default install paths for NoMachine, my vote is for the paths I’ve used (above). Its just a vote, unjustified 🙂