VMD-L Mailing List

From: Gianluca Interlandi (gianluca_at_u.washington.edu)
Date: Fri Dec 08 2023 - 16:17:17 CST

Hi again!

I finally installed a NVIDIA video card (GT 1030) and installed the NVIDIA
proprietary drivers (535.146.02). And you were right, even with the NVIDIA
proprietary drivers I still cannot open a remote openGL window. It's not a
big deal since I have been using `sshfs` and that has worked pretty well.

Thanks all for your feedback,

                            Gianluca

On Thu, 2 Nov 2023, Axel Kohlmeyer wrote:

> If I *have* to use VMD on a remote machine, repeatedly (or have to support people that need/want to
> do that), I download the Mesa sources and compile a software-rendering only version (so no DRI
> drivers and any of the other fancy stuff) and then use LD_PRELOAD to substitute the local OpenGL
> library on the remote machine with it. Of course, this is agonizingly slow, but it is completely
> independent from GLX support since it used plain X11 primitives only. On our HPC clusters we have
> such a feature for people that need to use OpenGL accelerated software from remote. There were times
> (20+ years ago) when this was the only option to run VMD and it was linked to Mesa directly, but
> fortunately, those times are long gone.
>
> On Thu, Nov 2, 2023 at 5:46?PM Gianluca Interlandi <gianluca_at_u.washington.edu> wrote:
> Hi Axel!
>
> Thanks for your reply.
>
> > This rarely works unless you have identical OS, X11, and OpenGL software
> > on both sides.
>
> So you are saying that having the same GPU with proprietary drivers may
> not help if everything else is different.
>
> > It most often breaks when the remote machine has more recent libraries
> > than the local one.
>
> In my situation it's the other way around, the local machine has more
> recent libraries.
>
> > So moving the data and running VMD locally is often the best solution.
>
> I found `sshfs` to be practical. Only caveat is that sshfs truncates the
> time stamp to the seconds, i.e.,
>
> Folder mounted with sshfs:
> > ls -lt --full-time test.trj
> 14:00:25.000000000 -0700 test.trj
>
> ssh to the remore machine:
> > ls -lt --full-time test.trj
> 14:00:25.720541063 -0700
>
> So I only use sshfs to read the trajectories locally with VMD but if I
> need to write new files I do it on the remote machine after logging in
> with ssh. Running VMD remotely is useful for example when performing an
> analysis that can only be done when VMD is in graphics mode. This used to
> be the case for example with parseFEP, but newer implementations of VMD
> allow parseFEP to be run through a script in text mode.
>
> Best,
>
>     Gianluca
>
> On Thu, 2 Nov 2023, Axel Kohlmeyer wrote:
>
> > When using a remote desktop solution the OpenGL processing is done remotely and then
> > the resulting graphics captured and then "streamed" to the remote desktop client.
> > With GLX (i.e. OpenGL over X11) the situation is much more complex: the OpenGL
> > directives are encapsulated into X11 protocol and then unpacked on the local machine
> > and processed by the local graphics card. This is very fragile since the OpenGL
> > implementations on both ends of the connection need to be compatible. This rarely
> > works unless you have identical OS, X11, and OpenGL software on both sides. It most
> > often breaks when the remote machine has more recent libraries than the local one.
> >
> > So moving the data and running VMD locally is often the best solution.
> >
> > On Thu, Nov 2, 2023 at 4:41?PM Gianluca Interlandi <gianluca_at_u.washington.edu> wrote:
> >       Hi Diego,
> >
> >       Thanks for the reply. My question is also about the clients. Do the
> >       clients from which you connect also have NVIDIA with proprietary drivers
> >       installed? What I'm trying to figure out is whether remote openGL (VMD)
> >       works if both server and client have NVIDIA GPUs both with the
> >       proprietary
> >       drivers installed.
> >
> >       I am using VMD 1.8.7 and 1.9.2b21, both with the same result.
> >
> >       > Have you considered using a remote desktop solution instead of ssh?
> >       Over
> >       > here we use Nice DCV.
> >
> >       I am considering such an option. My understanding is that rdesktop works
> >       only to connect to a windows server. Is Nice DCV included in major
> >       distributions? Right now I am using `sshfs` to mount the remote directory
> >       and then start VMD locally.
> >
> >       Thanks!
> >
> >             Gianluca
> >
> >       On Thu, 2 Nov 2023, Diego Gomes wrote:
> >
> >       > Hi Gianluca, although I can't help with the X error, I can confirm the
> >       issue
> >       > with VMD.Can you please share which VMD version you are using?
> >       >
> >       > For me it works very well when I ssh -X to our head-node, which does
> >       not contain a
> >       > dedicated GPU, while using vmd 1.9.3 and 1.9.4a55. It fails with vmd
> >       1.9.4a58.
> >       > For any other machine the OpenGl windows fails either using "ssh -X" or
> >       "ssh -Y", vmd
> >       > 1.9.3, 1.9.4a55, and 1.9.4a58.
> >       > XRequest.130: BadRequest (invalid request code or no such operation)
> >       0x3600090 
> >       >
> >       > Machines are Ubuntu 2022.04.2, the ones with GPU have Nvidia Driver
> >       Version:
> >       > 530.30.02  CUDA Version: 12.1.
> >       >
> >       > Have you considered using a remote desktop solution instead of ssh?
> >       Over here we use
> >       > Nice DCV.
> >       >
> >       > Friendly regards,
> >       > Diego.
> >       >
> >       >
> >
> >       -----------------------------------------------------
> >       Gianluca Interlandi, PhD gianluca_at_u.washington.edu
> >                            +1 (206) 685 4435
> >                        
>   https://urldefense.com/v3/__http://gianluca.today/__;!!DZ3fjg!9dM36IdGRdmFnQn3PC8kw
> >       CuiFnDlXnfxOoFtpV827RN1u-glDiSMsPpDwciy4mJV-EBmLoJs4veR78k8SzpAuMU3FOU$
> >
> >       Department of Bioengineering
> >       University of Washington, Seattle WA U.S.A.
> >       -----------------------------------------------------
> >
> >
> >
> > --
> > Dr. Axel Kohlmeyer  akohlmey_at_gmail.com  https://urldefense.com/v3/__http://goo.gl/1wk0__;!!DZ3fjg!4sJDJG0IbITQVdXRdiFzoeghS52rcPgIemXHIi_PwdlQIKiSjarxJDjAc9CKGGwols84k3WGim0bf0UeDcUnyR6V4fk$
> > College of Science & Technology, Temple University, Philadelphia PA, USA
> >
> >
>
> -----------------------------------------------------
> Gianluca Interlandi, PhD gianluca_at_u.washington.edu
>                      +1 (206) 685 4435
>                      https://urldefense.com/v3/__http://gianluca.today/__;!!DZ3fjg!4sJDJG0IbITQVdXRdiFzoeghS52rcPgIemXHIi_PwdlQIKiSjarxJDjAc9CKGGwols84k3WGim0bf0UeDcUnFaAuF5Q$
>
> Department of Bioengineering
> University of Washington, Seattle WA U.S.A.
> -----------------------------------------------------
>
>
>
> --
> Dr. Axel Kohlmeyer  akohlmey_at_gmail.com  https://urldefense.com/v3/__http://goo.gl/1wk0__;!!DZ3fjg!4sJDJG0IbITQVdXRdiFzoeghS52rcPgIemXHIi_PwdlQIKiSjarxJDjAc9CKGGwols84k3WGim0bf0UeDcUnyR6V4fk$
> College of Science & Technology, Temple University, Philadelphia PA, USA
>
>

-----------------------------------------------------
Gianluca Interlandi, PhD gianluca_at_u.washington.edu
                     +1 (206) 685 4435
                     https://urldefense.com/v3/__http://gianluca.today/__;!!DZ3fjg!4sJDJG0IbITQVdXRdiFzoeghS52rcPgIemXHIi_PwdlQIKiSjarxJDjAc9CKGGwols84k3WGim0bf0UeDcUnFaAuF5Q$

Department of Bioengineering
University of Washington, Seattle WA U.S.A.
-----------------------------------------------------