Menu
▾
▴
vizstack-users — Mailing list for everyone who uses VizStack
You can subscribe to this list here.
| 2009 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
(1) |
Dec
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2010 |
Jan
|
Feb
(3) |
Mar
(8) |
Apr
(26) |
May
(41) |
Jun
(3) |
Jul
(19) |
Aug
|
Sep
|
Oct
|
Nov
(2) |
Dec
|
| 2011 |
Jan
(17) |
Feb
(5) |
Mar
(1) |
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2012 |
Jan
|
Feb
|
Mar
(1) |
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|
From: Carlile, K. <car...@ja...> - 2012-03-13 16:24:59
|
2 quick questions: 1. Is it possible to run the master machine as a VM, and if so, does that VM need to be running the same distro/version as the nodes? 2. Does the master need NVidia drivers installed if it does not contain any GPUs? Thanks, Ken |
|
From: Kumar, S. <shr...@hp...> - 2011-03-01 08:42:00
|
With the newer nvidia series 270 drivers, the configuration script would fail on systems equipped with a single GPU. The 270 series drivers now store information about GPUs in /proc/driver/nvidia/gpus/[0-9]/information. The earlier drivers had it in /proc/driver/nvidia/cards/[0-9]. This issue has been fixed in SVN trunk. For more information, please refer SF Bug ID 3195935 @ https://sourceforge.net/tracker/?func=detail&aid=3195935&group_id=273582&atid=1162783 HTH -- Shree |
|
From: Kumar, S. <shr...@hp...> - 2011-02-11 09:10:23
|
Hi Peter, Thanks for the update. I have made changes to "trunk/src/vs-Xv.cpp". You'll have to update, rebuild the RPM & propagate the RPM to all nodes. I am _hoping_ that this change will fix the issue. Do try this out & let me know if the problem gets fixed. If this does not fix the problem, then I'll have to try other ideas. Cheers -- Shree -----Original Message----- From: Pet...@cs... [mailto:Pet...@cs...] Sent: Friday, February 11, 2011 11:53 AM To: viz...@li... Subject: Re: [vizstack-users] Vizstack issue - leaving behind /tmp/.X... files Hi Shree, Thanks for your quick reply. I should have mentioned that when the problem occurs both the VNC server and your scripts have terminated. All that is required to have things working again is to delete the files left behind in /tmp. I agree with you that the first approach seems the best way to go. Cheers, Peter From: Kumar, Shree [shr...@hp...] Sent: Thursday, 10 February 2011 8:20 PM To: viz...@li... Subject: Re: [vizstack-users] Vizstack issue - leaving behind /tmp/.X... files Hi Peter, Please check this : When the lock files are left behind, are any "vs-Xv" processes active ? Are the VNC servers still running ? This is a problematic case really. The TurboVNC server is currently started via a wrapper, named "vs-Xv". The wrapper attempts to clean up the lock files when the VNC server exits. I can think of two solutions 1. Termination of the wrapper is not handled at the moment. If I catch this, then I can enforce deep cleanup 2. vs-Xv can cleanup the log files before I start the VNC server. Since the log files can potentially belong to another user, vs-Xv will need to be made SUID root. I feel inclined to try approach #1, since approach #2 has security implications. What do you think ? Cheers -- Shree -----Original Message----- From: Pet...@cs... [mailto:Pet...@cs...] Sent: Thursday, February 10, 2011 11:56 AM To: viz...@li... Subject: [vizstack-users] Vizstack issue - leaving behind /tmp/.X... files Hi, I've been seeing an issue with Vizstack where at times /.X*-lock and /tmp/.X11-unix/X* files are being left behind from VNC sessions. The files relating to the 3D X servers are always cleaned up, the files in question are from VNC sessions. This seems to have only occurred when a user has terminated their session by closing the SSH application (usually Putty on Windows) or shutdown their PC/logged out with a session still active rather than pressing cntl-c to take them back to the prompt. The job has always been removed from the SLURM queue in these cases. This problem is intermittent although it has been happening quite frequently. Having these lock files left behind causes quite a problem of course as Vizstack will often want to use that display number for the next VNC session on that particular node. I'm using a late SVN version of Vizstack with TurboVNC 1.0 on Ubuntu 10.04. Any ideas? Cheers, Peter Peter Tyson CSIRO IM&T - Advanced Scientific Computing Gate 5 Normanby Road Clayton Vic 3168 Ph +61 3 9545 2021 ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users |
|
From: <Pet...@cs...> - 2011-02-11 06:23:35
|
Hi Shree, Thanks for your quick reply. I should have mentioned that when the problem occurs both the VNC server and your scripts have terminated. All that is required to have things working again is to delete the files left behind in /tmp. I agree with you that the first approach seems the best way to go. Cheers, Peter From: Kumar, Shree [shr...@hp...] Sent: Thursday, 10 February 2011 8:20 PM To: viz...@li... Subject: Re: [vizstack-users] Vizstack issue - leaving behind /tmp/.X... files Hi Peter, Please check this : When the lock files are left behind, are any "vs-Xv" processes active ? Are the VNC servers still running ? This is a problematic case really. The TurboVNC server is currently started via a wrapper, named "vs-Xv". The wrapper attempts to clean up the lock files when the VNC server exits. I can think of two solutions 1. Termination of the wrapper is not handled at the moment. If I catch this, then I can enforce deep cleanup 2. vs-Xv can cleanup the log files before I start the VNC server. Since the log files can potentially belong to another user, vs-Xv will need to be made SUID root. I feel inclined to try approach #1, since approach #2 has security implications. What do you think ? Cheers -- Shree -----Original Message----- From: Pet...@cs... [mailto:Pet...@cs...] Sent: Thursday, February 10, 2011 11:56 AM To: viz...@li... Subject: [vizstack-users] Vizstack issue - leaving behind /tmp/.X... files Hi, I've been seeing an issue with Vizstack where at times /.X*-lock and /tmp/.X11-unix/X* files are being left behind from VNC sessions. The files relating to the 3D X servers are always cleaned up, the files in question are from VNC sessions. This seems to have only occurred when a user has terminated their session by closing the SSH application (usually Putty on Windows) or shutdown their PC/logged out with a session still active rather than pressing cntl-c to take them back to the prompt. The job has always been removed from the SLURM queue in these cases. This problem is intermittent although it has been happening quite frequently. Having these lock files left behind causes quite a problem of course as Vizstack will often want to use that display number for the next VNC session on that particular node. I'm using a late SVN version of Vizstack with TurboVNC 1.0 on Ubuntu 10.04. Any ideas? Cheers, Peter Peter Tyson CSIRO IM&T - Advanced Scientific Computing Gate 5 Normanby Road Clayton Vic 3168 Ph +61 3 9545 2021 ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users |
|
From: Kumar, S. <shr...@hp...> - 2011-02-10 09:21:47
|
Hi Peter, Please check this : When the lock files are left behind, are any "vs-Xv" processes active ? Are the VNC servers still running ? This is a problematic case really. The TurboVNC server is currently started via a wrapper, named "vs-Xv". The wrapper attempts to clean up the lock files when the VNC server exits. I can think of two solutions 1. Termination of the wrapper is not handled at the moment. If I catch this, then I can enforce deep cleanup 2. vs-Xv can cleanup the log files before I start the VNC server. Since the log files can potentially belong to another user, vs-Xv will need to be made SUID root. I feel inclined to try approach #1, since approach #2 has security implications. What do you think ? Cheers -- Shree -----Original Message----- From: Pet...@cs... [mailto:Pet...@cs...] Sent: Thursday, February 10, 2011 11:56 AM To: viz...@li... Subject: [vizstack-users] Vizstack issue - leaving behind /tmp/.X... files Hi, I've been seeing an issue with Vizstack where at times /.X*-lock and /tmp/.X11-unix/X* files are being left behind from VNC sessions. The files relating to the 3D X servers are always cleaned up, the files in question are from VNC sessions. This seems to have only occurred when a user has terminated their session by closing the SSH application (usually Putty on Windows) or shutdown their PC/logged out with a session still active rather than pressing cntl-c to take them back to the prompt. The job has always been removed from the SLURM queue in these cases. This problem is intermittent although it has been happening quite frequently. Having these lock files left behind causes quite a problem of course as Vizstack will often want to use that display number for the next VNC session on that particular node. I'm using a late SVN version of Vizstack with TurboVNC 1.0 on Ubuntu 10.04. Any ideas? Cheers, Peter Peter Tyson CSIRO IM&T - Advanced Scientific Computing Gate 5 Normanby Road Clayton Vic 3168 Ph +61 3 9545 2021 ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users |
|
From: <Pet...@cs...> - 2011-02-10 06:26:54
|
Hi, I've been seeing an issue with Vizstack where at times /.X*-lock and /tmp/.X11-unix/X* files are being left behind from VNC sessions. The files relating to the 3D X servers are always cleaned up, the files in question are from VNC sessions. This seems to have only occurred when a user has terminated their session by closing the SSH application (usually Putty on Windows) or shutdown their PC/logged out with a session still active rather than pressing cntl-c to take them back to the prompt. The job has always been removed from the SLURM queue in these cases. This problem is intermittent although it has been happening quite frequently. Having these lock files left behind causes quite a problem of course as Vizstack will often want to use that display number for the next VNC session on that particular node. I'm using a late SVN version of Vizstack with TurboVNC 1.0 on Ubuntu 10.04. Any ideas? Cheers, Peter Peter Tyson CSIRO IM&T - Advanced Scientific Computing Gate 5 Normanby Road Clayton Vic 3168 Ph +61 3 9545 2021 |
|
From: Kumar, S. <shr...@hp...> - 2011-02-07 04:50:12
|
Hi Cristi,
It's been a while, and I haven't heard from you on this. Did you get around to solving this problem ? If you need more help, then let me know.
Regards
-- Shree
-----Original Message-----
From: Kumar, Shree
Sent: Thursday, January 27, 2011 10:40 AM
To: viz...@li...
Subject: Re: [vizstack-users] Configuring VizStack - using TurboVNC - Problem
Hi Cristi,
This is indeed possible to do by modifying viz-tvnc (it's written in Python). There are issues to consider.
First, you need to figure out the port number to use on the master public IP address. Your end users could execute the viz-tvnc script in any order, and each would use a port on the master IP. Their usage of ports must not clash, and you need to ensure this. How will you ensure this ? If every user is allowed only one session, then this becomes easy. For each user, you can predetermine a port number to be used for forwarding for that user. This approach is simple to implement, and has the benefit of avoiding the race condition mentioned below. If you allow one user only one session, then this is the way to go !
Second, sessions could be started and killed. Setting up port forwarding should not cause race conditions in the script.
What's your view on these issues ?
Cheers
-- Shree
-----Original Message-----
From: Cristi Mocan [mailto:Cri...@ne...]
Sent: Wednesday, January 26, 2011 4:42 PM
To: viz...@li...
Subject: Re: [vizstack-users] Configuring VizStack - using TurboVNC - Problem
Hi Shree,
Related to your answer, I have to say that I need the second way - Port forwarding.
I have used port forwarding like in your example:
ssh -N -L <Master Public IP Address>:5901:192.168.0.2:5901 192.168.0.2
So, after I have started viz-tvnc on the master node and see the command line interface in the terminal ("A desktop has been started for you at
node2:1 ...." - I started another terminal on the master node and ran this ssh command with <Master Public IP Address>:5901. After entering the password, I could connect remotely using the TurboVNC Viewer.
Can you help me with this issue in order to set this ssh command in viz-tvnc (Python) file. Is there any possibility to have this port forwarding resolved dynamically in viz-tvnc ? For now I have just one user.
Thank you in advance !
Regards,
Cristi
------------------------------------------------------------------------------
Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)!
Finally, a world-class log management solution at an even better price-free!
Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY!
http://p.sf.net/sfu/arcsight-sfd2d
_______________________________________________
vizstack-users mailing list
viz...@li...
https://lists.sourceforge.net/lists/listinfo/vizstack-users
------------------------------------------------------------------------------
Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)!
Finally, a world-class log management solution at an even better price-free!
Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY!
http://p.sf.net/sfu/arcsight-sfd2d
_______________________________________________
vizstack-users mailing list
viz...@li...
https://lists.sourceforge.net/lists/listinfo/vizstack-users
|
|
From: Kumar, S. <shr...@hp...> - 2011-01-27 05:11:01
|
Hi Cristi,
This is indeed possible to do by modifying viz-tvnc (it's written in Python). There are issues to consider.
First, you need to figure out the port number to use on the master public IP address. Your end users could execute the viz-tvnc script in any order, and each would use a port on the master IP. Their usage of ports must not clash, and you need to ensure this. How will you ensure this ? If every user is allowed only one session, then this becomes easy. For each user, you can predetermine a port number to be used for forwarding for that user. This approach is simple to implement, and has the benefit of avoiding the race condition mentioned below. If you allow one user only one session, then this is the way to go !
Second, sessions could be started and killed. Setting up port forwarding should not cause race conditions in the script.
What's your view on these issues ?
Cheers
-- Shree
-----Original Message-----
From: Cristi Mocan [mailto:Cri...@ne...]
Sent: Wednesday, January 26, 2011 4:42 PM
To: viz...@li...
Subject: Re: [vizstack-users] Configuring VizStack - using TurboVNC - Problem
Hi Shree,
Related to your answer, I have to say that I need the second way - Port forwarding.
I have used port forwarding like in your example:
ssh -N -L <Master Public IP Address>:5901:192.168.0.2:5901 192.168.0.2
So, after I have started viz-tvnc on the master node and see the command line interface in the terminal ("A desktop has been started for you at
node2:1 ...." - I started another terminal on the master node and ran this ssh command with <Master Public IP Address>:5901. After entering the password, I could connect remotely using the TurboVNC Viewer.
Can you help me with this issue in order to set this ssh command in viz-tvnc (Python) file. Is there any possibility to have this port forwarding resolved dynamically in viz-tvnc ? For now I have just one user.
Thank you in advance !
Regards,
Cristi
------------------------------------------------------------------------------
Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)!
Finally, a world-class log management solution at an even better price-free!
Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY!
http://p.sf.net/sfu/arcsight-sfd2d
_______________________________________________
vizstack-users mailing list
viz...@li...
https://lists.sourceforge.net/lists/listinfo/vizstack-users
|
|
From: Cristi M. <Cri...@ne...> - 2011-01-26 11:12:11
|
Hi Shree,
Related to your answer, I have to say that I need the second way - Port
forwarding.
I have used port forwarding like in your example:
ssh -N -L <Master Public IP Address>:5901:192.168.0.2:5901 192.168.0.2
So, after I have started viz-tvnc on the master node and see the command
line interface in the terminal ("A desktop has been started for you at
node2:1 ...." - I started another terminal on the master node and ran
this ssh command with <Master Public IP Address>:5901. After entering the
password, I could connect remotely using the TurboVNC Viewer.
Can you help me with this issue in order to set this ssh command in
viz-tvnc (Python) file. Is there any possibility to have this port
forwarding resolved dynamically in viz-tvnc ? For now I have just one
user.
Thank you in advance !
Regards,
Cristi
|
|
From: Kumar, S. <shr...@hp...> - 2011-01-23 15:15:23
|
Hi Cristi, Good to see things are working. I see two ways around your current problem : - Use the reverse connection feature, accessible using "-c". Use this if you have a public IP address for the system where the TurboVNC client will be run. You'll have to start the TurboVNC client in "listening mode" before you run the script. - Use port forwarding. Setup an SSH tunnel from your master node to the TurboVNC server session on the rendering nodes. TurboVNC servers use port number (5900+server number). You can do this manually, or by modifying the viz-tvnc script. You'd use the "-N" and "-L" options. E.g., if your TurboVNC session was allocated on 192.168.0.2:1, you could setup an SSH tunnel as follows: ssh -N -L <Master Public IP Address>:10000:192.168.0.2:5901 192.168.0.2 Finding a source port to use can be a pain, depending on how many users you have. Regards -- Shree -----Original Message----- From: Cristi Mocan [mailto:Cri...@ne...] Sent: Friday, January 21, 2011 7:52 PM To: viz...@li... Subject: Re: [vizstack-users] Configuring VizStack - using TurboVNC - Problem] Hi Shree, Related to my last post, I have to say: -Thank you very much for your help Shree ! -Mission accomplished from this point of view ! Now I have to do one more important thing: First of all I would like to present my hardware configuration. I have a cluster that consists: - node0 (server - master node -SSM) - node1, node2,..,nodeX (rendering nodes) I have to mention that my master node is a Virtual Machine and does not have a GPU. The rendering nodes have NVIDIA GTS 8800. All the nodes have OS: Ubuntu Server 10.04 Node0 has two network interfaces: -eth0: Internet connection - Public Real IP -eth1: LAN connection - Static IP: 192.168.0.1 Using Vizstack with viz-tvnc command - the VNC server will start different desktop sessions on node2,node3... 192.168.0.2:1 192.168.0.2:2 192.168.0.3:1 192.168.0.4....etc How should I setup the visualization cluster in order to access these desktop sessions remotely using(through) the Master(Server) Public IP ?? Is it possible to make port forwarding ? I am sure you can give me the right solution for this issue. Thank you very much for your help ! Cheers, Cristi ------------------------------------------------------------------------------ Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)! Finally, a world-class log management solution at an even better price-free! Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY! http://p.sf.net/sfu/arcsight-sfd2d _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users |
|
From: Cristi M. <Cri...@ne...> - 2011-01-21 14:22:14
|
Hi Shree, Related to my last post, I have to say: -Thank you very much for your help Shree ! -Mission accomplished from this point of view ! Now I have to do one more important thing: First of all I would like to present my hardware configuration. I have a cluster that consists: - node0 (server - master node -SSM) - node1, node2,..,nodeX (rendering nodes) I have to mention that my master node is a Virtual Machine and does not have a GPU. The rendering nodes have NVIDIA GTS 8800. All the nodes have OS: Ubuntu Server 10.04 Node0 has two network interfaces: -eth0: Internet connection - Public Real IP -eth1: LAN connection - Static IP: 192.168.0.1 Using Vizstack with viz-tvnc command - the VNC server will start different desktop sessions on node2,node3... 192.168.0.2:1 192.168.0.2:2 192.168.0.3:1 192.168.0.4....etc How should I setup the visualization cluster in order to access these desktop sessions remotely using(through) the Master(Server) Public IP ?? Is it possible to make port forwarding ? I am sure you can give me the right solution for this issue. Thank you very much for your help ! Cheers, Cristi |
|
From: Kumar, S. <shr...@hp...> - 2011-01-21 10:25:55
|
Hi Cristi,
I feel we are closed to solving this. What's the current permission on ~/.vnc ?
My advice is that you fix the permissions on the ~/.vnc directory. E.g., on my system, ~/.vnc has permissions
drwx------ 2 shree shree 4096 Jan 21 12:34 .vnc
Regards
-- Shree
-----Original Message-----
From: Cristi Mocan [mailto:Cri...@ne...]
Sent: Friday, January 21, 2011 3:47 PM
To: viz...@li...
Subject: Re: [vizstack-users] Configuring VizStack - TurboVNC server failed to start - Problem
From: Kumar, Shree <shreekumar@hp...> - 2011-01-21 05:10
Hi Cristi,
Bad guess! Ok - so now we have to look deeper. Thanks for the additional information.
>>>>>> Hi Shree ! Thank you for the replies.
VizStack runs a daemon on port 50000 by default. The scripts from the other nodes connect to this daemon. Please ensure that your virtual machine is not blocking this connection. If this port is being blocked, then open it and then retry the viz-tvnc script.
>>>>>>> The port is not blocked.
Next, try running /opt/vizstack/sbin/vs-test-gpus. This should return successfully.
>>>>>>> Test PASSED
If all these do not fix the problem, then make a copy of /opt/vizstack/bin/viz-tvnc. Edit the file. Locate the line
tvncProc = sched.run(cmd, errFile = open("/dev/null","w"))
And edit it to see the error messages
tvncProc = sched.run(cmd)
Run the edited copy of viz-tvnc. You will see error messages. Pass them on to me to have a look.
>>>>>>> After I did the step above I found the problem ! Good Point !
>>>>>>>
>>>>>>> >viz-tvnc
>>>>>>> >Starting Desktop with resolution 1280x1024
>>>>>>> >vs-vncserver: Wrong type or access mode of /home/cristi/.vnc.
>>>>>>> >TurboVNC server failed to start.
>>>>>>>
>>>>>>> So, the proble came from this vs-vncserver file: line 237-238
>>>>>>>
>>>>>>>
if (!-d _ || !-o _ || ($vncUserDirUnderTmp && ($mode & 0777) != 0700)) {
die "$prog: Wrong type or access mode of $vncUserDir.\n"; }
>>>>>>> What I should do in order to fix this attributes-rights...
>>>>>>> What is your advice ?
>>>>>>> When I tried to run a standalone config directly on the node, it
>>>>>>> returns the same message (of course).
>>>>>>>
>>>>>>> Thank you in advance !
>>>>>>>
>>>>>>>Cheers,
>>>>>>>Cristi
Regards
-- Shree
------------------------------------------------------------------------------
Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)!
Finally, a world-class log management solution at an even better price-free!
Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY!
http://p.sf.net/sfu/arcsight-sfd2d
_______________________________________________
vizstack-users mailing list
viz...@li...
https://lists.sourceforge.net/lists/listinfo/vizstack-users
|
|
From: Cristi M. <Cri...@ne...> - 2011-01-21 10:17:31
|
From: Kumar, Shree <shreekumar@hp...> - 2011-01-21 05:10
Hi Cristi,
Bad guess! Ok - so now we have to look deeper. Thanks for the additional
information.
>>>>>> Hi Shree ! Thank you for the replies.
VizStack runs a daemon on port 50000 by default. The scripts from the
other nodes connect to this daemon. Please ensure that your virtual
machine is not blocking this connection. If this port is being blocked,
then open it and then retry the viz-tvnc script.
>>>>>>> The port is not blocked.
Next, try running /opt/vizstack/sbin/vs-test-gpus. This should return
successfully.
>>>>>>> Test PASSED
If all these do not fix the problem, then make a copy of
/opt/vizstack/bin/viz-tvnc. Edit the file. Locate the line
tvncProc = sched.run(cmd, errFile = open("/dev/null","w"))
And edit it to see the error messages
tvncProc = sched.run(cmd)
Run the edited copy of viz-tvnc. You will see error messages. Pass them on
to me to have a look.
>>>>>>> After I did the step above I found the problem ! Good Point !
>>>>>>>
>>>>>>> >viz-tvnc
>>>>>>> >Starting Desktop with resolution 1280x1024
>>>>>>> >vs-vncserver: Wrong type or access mode of /home/cristi/.vnc.
>>>>>>> >TurboVNC server failed to start.
>>>>>>>
>>>>>>> So, the proble came from this vs-vncserver file: line 237-238
>>>>>>>
>>>>>>>
if (!-d _ || !-o _ || ($vncUserDirUnderTmp && ($mode & 0777) != 0700)) {
die "$prog: Wrong type or access mode of $vncUserDir.\n";
}
>>>>>>> What I should do in order to fix this attributes-rights...
>>>>>>> What is your advice ?
>>>>>>> When I tried to run a standalone config directly on the node, it
>>>>>>> returns the same message (of course).
>>>>>>>
>>>>>>> Thank you in advance !
>>>>>>>
>>>>>>>Cheers,
>>>>>>>Cristi
Regards
-- Shree
|
|
From: Kumar, S. <shr...@hp...> - 2011-01-21 05:10:25
|
Hi Cristi,
Bad guess! Ok - so now we have to look deeper. Thanks for the additional information.
VizStack runs a daemon on port 50000 by default. The scripts from the other nodes connect to this daemon. Please ensure that your virtual machine is not blocking this connection. If this port is being blocked, then open it and then retry the viz-tvnc script.
Next, try running /opt/vizstack/sbin/vs-test-gpus. This should return successfully.
If all these do not fix the problem, then make a copy of /opt/vizstack/bin/viz-tvnc. Edit the file. Locate the line
tvncProc = sched.run(cmd, errFile = open("/dev/null","w"))
And edit it to see the error messages
tvncProc = sched.run(cmd)
Run the edited copy of viz-tvnc. You will see error messages. Pass them on to me to have a look.
Regards
-- Shree
-----Original Message-----
From: Cristi Mocan [mailto:Cri...@ne...]
Sent: Thursday, January 20, 2011 9:56 PM
To: viz...@li...
Subject: Re: [vizstack-users] Configuring VizStack - TurboVNC server failed to start - Problem
First of all I would like to present my hardware configuration.
I have a cluster that consists:
- node0 (server - master node -SSM)
- node1, node2 (rendering nodes)
I have to mention that my master node is a Virtual Machine and does not have a GPU. The rendering nodes have NVIDIA GTS 8800. All the nodes have
OS: Ubuntu Server 10.04
I would like to add that I had tested the VNC server (classical way) and is working fine on each node (node0, node1, node2).
/opt/TurboVNC/bin/vncserver
I was asking for a password on each node. On every node I have the same user and the home directory contains the same /.vnc folder with the same passwd file. So, I have the same user & VNC password everywhere in the cluster.
I don't know why is not working using Vizstack....
After I run the config:
/opt/vizstack/sbin/vs-configure-system -s slurm node[1-2] and start the SSM: /opt/vizstack/sbin/vs-ssm start
When I run:
viz-tvnc
The result is the same:
Starting Desktop with resolution 1280x1024 TurboVNC server failed to start.
I receive the same result even if I try with:
/opt/vizstack/sbin/vs-configure-system -s slurm node1 or /opt/vizstack/sbin/vs-configure-system -s slurm node2
I received the same result if I tried the standalone config from node1 or
node2:
/opt/vizstack/sbin/vs-configure-standalone
I would appreciate if you can give me some more information about how I should setup Vizstack or TurboVNC user pass...
Thank you in advance !
Cheers,
Cristi
----------------------------------------------------------------------------
From: Kumar, Shree <shreekumar@hp...> - 2011-01-20 04:48
Hi Cristi,
Looks like a commonn cluster setup issue to me.
>From your earlier email, I can see that VizStack is managing GPUs on
>two
systems cgis[1-2]. When you run viz-tvnc, it tries to start a TurboVNC server on one of those systems. If a TurboVNC server cannot be started on those nodes, then the script will fail.
You have already setup a password on your local machine (the machine where you ran the configure scrpt - cgis0, perhaps), so the server is starting up just fine.
Interestingly, startup can fail on a per-user basis. If a user has not assigned a vnc password, then the server will not start. If you try to start the TurboVNC server on any of the nodes cgis[1-2], you will find that it prompts you for a password. Can you confirm that this is the issue ?
If you find that this is indeed the issue, then you will have to ensure that every user who wants to use TurboVNC will have to set a password on every machine. If you have shared home directories, then they need to set the password atleast once.
Regards
-- Shree
------------------------------------------------------------------------------
Protect Your Site and Customers from Malware Attacks Learn about various malware tactics and how to avoid them. Understand malware threats, the impact they can have on your business, and how you can protect your company and customers by using code signing.
http://p.sf.net/sfu/oracle-sfdevnl
_______________________________________________
vizstack-users mailing list
viz...@li...
https://lists.sourceforge.net/lists/listinfo/vizstack-users
|
|
From: Cristi M. <Cri...@ne...> - 2011-01-20 16:26:14
|
First of all I would like to present my hardware configuration. I have a cluster that consists: - node0 (server - master node -SSM) - node1, node2 (rendering nodes) I have to mention that my master node is a Virtual Machine and does not have a GPU. The rendering nodes have NVIDIA GTS 8800. All the nodes have OS: Ubuntu Server 10.04 I would like to add that I had tested the VNC server (classical way) and is working fine on each node (node0, node1, node2). /opt/TurboVNC/bin/vncserver I was asking for a password on each node. On every node I have the same user and the home directory contains the same /.vnc folder with the same passwd file. So, I have the same user & VNC password everywhere in the cluster. I don't know why is not working using Vizstack.... After I run the config: /opt/vizstack/sbin/vs-configure-system -s slurm node[1-2] and start the SSM: /opt/vizstack/sbin/vs-ssm start When I run: viz-tvnc The result is the same: Starting Desktop with resolution 1280x1024 TurboVNC server failed to start. I receive the same result even if I try with: /opt/vizstack/sbin/vs-configure-system -s slurm node1 or /opt/vizstack/sbin/vs-configure-system -s slurm node2 I received the same result if I tried the standalone config from node1 or node2: /opt/vizstack/sbin/vs-configure-standalone I would appreciate if you can give me some more information about how I should setup Vizstack or TurboVNC user pass... Thank you in advance ! Cheers, Cristi ---------------------------------------------------------------------------- From: Kumar, Shree <shreekumar@hp...> - 2011-01-20 04:48 Hi Cristi, Looks like a commonn cluster setup issue to me. >From your earlier email, I can see that VizStack is managing GPUs on two systems cgis[1-2]. When you run viz-tvnc, it tries to start a TurboVNC server on one of those systems. If a TurboVNC server cannot be started on those nodes, then the script will fail. You have already setup a password on your local machine (the machine where you ran the configure scrpt - cgis0, perhaps), so the server is starting up just fine. Interestingly, startup can fail on a per-user basis. If a user has not assigned a vnc password, then the server will not start. If you try to start the TurboVNC server on any of the nodes cgis[1-2], you will find that it prompts you for a password. Can you confirm that this is the issue ? If you find that this is indeed the issue, then you will have to ensure that every user who wants to use TurboVNC will have to set a password on every machine. If you have shared home directories, then they need to set the password atleast once. Regards -- Shree |
|
From: Kumar, S. <shr...@hp...> - 2011-01-20 04:48:59
|
Hi Cristi, Looks like a commonn cluster setup issue to me. >From your earlier email, I can see that VizStack is managing GPUs on two systems cgis[1-2]. When you run viz-tvnc, it tries to start a TurboVNC server on one of those systems. If a TurboVNC server cannot be started on those nodes, then the script will fail. You have already setup a password on your local machine (the machine where you ran the configure scrpt - cgis0, perhaps), so the server is starting up just fine. Interestingly, startup can fail on a per-user basis. If a user has not assigned a vnc password, then the server will not start. If you try to start the TurboVNC server on any of the nodes cgis[1-2], you will find that it prompts you for a password. Can you confirm that this is the issue ? If you find that this is indeed the issue, then you will have to ensure that every user who wants to use TurboVNC will have to set a password on every machine. If you have shared home directories, then they need to set the password atleast once. Regards -- Shree -----Original Message----- From: Cristi Mocan [mailto:Cri...@ne...] Sent: Thursday, January 20, 2011 5:57 AM To: viz...@li... Subject: Re: [vizstack-users] Configuring VizStack - TurboVNC server failed to start - Problem I would like to add that I had tested the VNC server (classical way)and is working fine. EX: /opt/TurboVNC/bin/vncserver New 'X' desktop is my_server:1 On the client machine, I started the TurboVNC Viewer. A TurboVNC desktop window appeared on the client machine (port 5901). This window contains a virtual X server with which I can interact. I don't know why is not working in Vizstack.... Thank you in advance ! Regards, Cristi ------------------------------------------------------------------------------ Protect Your Site and Customers from Malware Attacks Learn about various malware tactics and how to avoid them. Understand malware threats, the impact they can have on your business, and how you can protect your company and customers by using code signing. http://p.sf.net/sfu/oracle-sfdevnl _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users |
|
From: Cristi M. <Cri...@ne...> - 2011-01-20 00:26:44
|
I would like to add that I had tested the VNC server (classical way)and is working fine. EX: /opt/TurboVNC/bin/vncserver New 'X' desktop is my_server:1 On the client machine, I started the TurboVNC Viewer. A TurboVNC desktop window appeared on the client machine (port 5901). This window contains a virtual X server with which I can interact. I don't know why is not working in Vizstack.... Thank you in advance ! Regards, Cristi |
|
From: Cristi M. <Cri...@ne...> - 2011-01-19 13:48:34
|
Hello again,
First of all I would like to say - Thank You Shree ! - for your notes
about the packages that need to be installed manually for Ubuntu Server !
Now I have the following use case:
On the master node (node0) I run the following
/opt/vizstack/sbin/vs-configure-system -s slurm cgis[1-2]
Everything is working fine here.
Result:
--------------------------------------------------------------------------
The summary of this configuration is :
SSM is running at host 'cgis0', port '50000'
- 2 nodes will be managed by the SSM
- 2 GPU(s) are available
- 1 GPU(s) are configured for shared usage
- scheduler used will be 'slurm'
- see file '/etc/vizstack/node_config.xml' for node configuration &
scheduler details
To activate this configuration, you need to start the SSM
# /opt/vizstack/sbin/vs-ssm start
One or more files were backed up during this configuration.
All the original files have been backed up with a name suffix
'.Wed_Jan_19_12:53:39_2011'
---------------------------------------------------------------------------
After this I run /opt/vizstack/sbin/vs-ssm start and the SSM Status is OK.
Result:
VizStack SSM daemon is running with pid=23656
Then I run: viz-tvnc
Result:
Starting Desktop with resolution 1280x1024
TurboVNC server failed to start.
So, I guess the problem regarding this TurboVNC server is comming from:
except vsapi.VizError, e:
showError(isAutomation,"TurboVNC server failed to start"
I have installed the following packages on all the nodes:
- turbovnc_1.0_amd64.deb
- VirtualGL_2.2_amd64.deb
- turbojpeg_1.11.1_amd64.deb
Do you know why the TurboVNC server failed to start ?
I tried to run these commands on another node with GPU:
/opt/vizstack/sbin/vs-configure-standalone
/opt/vizstack/sbin/vs-ssm start
viz-tvnc
Result:
Starting Desktop with resolution 1280x1024
TurboVNC server failed to start.
Thank you in advance !
Regards,
Cristi
|
|
From: Kumar, S. <shr...@hp...> - 2011-01-18 04:51:49
|
Hi Cristi, Looks like you were right about some parts of the X server missing. You have an incomplete X server installation. I can think of one way to fix this: run "xinit" from the command prompt. The shell will tell you which package needs to be installed. Go on and install the package. After that, the vs-configure-standalone command would succeed. Getting a working installation from a barebones server installation of Ubuntu is quite some work ! You will keep hitting situations where binaries are not available. I have some notes about the packages that need to be installed manually for Ubuntu Server at this location: http://sourceforge.net/apps/trac/vizstack/browser/trunk/INSTALL Hope that helps -- Shree -----Original Message----- From: Cristi Mocan [mailto:Cri...@ne...] Sent: Monday, January 17, 2011 7:29 PM To: viz...@li... Subject: Re: [vizstack-users] Configuring VizStack - vs-generate-node-config Problem Hello again, Thank you for the solution: http://sourceforge.net/apps/trac/vizstack/changeset/339 If I applied the above mentioned change on one of the nodes (with GPU), I received this: Ex.: sudo /opt/vizstack/sbin/vs-configure-standalone Processing Node 'localhost'... Errors happened while trying to get the configuration of node 'localhost'. Reason: Traceback (most recent call last): File "/opt/vizstack/sbin/vs-generate-node-config", line 434, in <module> ret = runSystemTimed("xinit /opt/vizstack/bin/vs-get-limits %s -- -logverbose 6 -config %s"%(temp_result_file, output_path), timeout=30) File "/opt/vizstack/sbin/vs-generate-node-config", line 202, in runSystemTimed proc = subprocess.Popen(cmd_list, stdout=output, stderr=output) File "/usr/lib/python2.6/subprocess.py", line 633, in __init__ errread, errwrite) File "/usr/lib/python2.6/subprocess.py", line 1139, in _execute_child raise child_exception OSError: [Errno 2] No such file or directory Please fix the above errors & run this tool again Failed to configure standalone configuration Thank you in advance ! CRisti ------------------------------------------------------------------------------ Protect Your Site and Customers from Malware Attacks Learn about various malware tactics and how to avoid them. Understand malware threats, the impact they can have on your business, and how you can protect your company and customers by using code signing. http://p.sf.net/sfu/oracle-sfdevnl _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users |
|
From: Cristi M. <Cri...@ne...> - 2011-01-17 13:58:44
|
Hello again, Thank you for the solution: http://sourceforge.net/apps/trac/vizstack/changeset/339 If I applied the above mentioned change on one of the nodes (with GPU), I received this: Ex.: sudo /opt/vizstack/sbin/vs-configure-standalone Processing Node 'localhost'... Errors happened while trying to get the configuration of node 'localhost'. Reason: Traceback (most recent call last): File "/opt/vizstack/sbin/vs-generate-node-config", line 434, in <module> ret = runSystemTimed("xinit /opt/vizstack/bin/vs-get-limits %s -- -logverbose 6 -config %s"%(temp_result_file, output_path), timeout=30) File "/opt/vizstack/sbin/vs-generate-node-config", line 202, in runSystemTimed proc = subprocess.Popen(cmd_list, stdout=output, stderr=output) File "/usr/lib/python2.6/subprocess.py", line 633, in __init__ errread, errwrite) File "/usr/lib/python2.6/subprocess.py", line 1139, in _execute_child raise child_exception OSError: [Errno 2] No such file or directory Please fix the above errors & run this tool again Failed to configure standalone configuration Thank you in advance ! CRisti |
|
From: Kumar, S. <shr...@hp...> - 2011-01-17 04:43:48
|
Hi Cristi, I appreciate you trying VizStack. xserver-xorg package is necessary to use VizStack. I believe it is installed already on your system. The issue you see has been fixed, but is not yet part of a release. http://sourceforge.net/apps/trac/vizstack/changeset/339 If you apply the above mentioned change on all your node(s), then the configuration must proceed normally. Let me know if that does not fix it! Cheers -- Shree -----Original Message----- From: Cristi Mocan [mailto:Cri...@ne...] Sent: Friday, January 14, 2011 4:46 PM To: viz...@li... Subject: Re: [vizstack-users] Configuring VizStack - vs-generate-node-config Problem Hello again, Unfortunately it doesn't work like this - either with: # /opt/vizstack/sbin/vs-configure-system -s slurm node[1-4] or # /opt/vizstack/sbin/vs-configure-standalone - on the master node(with GPU) I received the same result: > Processing Node 'node5'... > Errors happened while trying to get the configuration of node 'node5'. > Reason: > > Traceback (most recent call last): > File "/opt/vizstack/sbin/vs-generate-node-config", line 304, in <module> > gpuInfo = parse_gpuInfo(temp_xconfig) > File "/opt/vizstack/sbin/vs-generate-node-config", line 168, in > parse_gpuInfo > busId = "PCI:%d:%02d:%01d"%(int(busIdParts[0], 16), > int(busIdParts[1], 16), int(busIdParts[2], 16)) > IndexError: list index out of range > srun: error: node1: task 0: Exited with exit code 1 > > > Please fix the above errors & run this tool again Do you have other solutions ? I am using vistack_1.1-3_ubuntu9.10.amb64.deb Thank you in advance. Cristi --------------------------------------------------------------------------- From: Manjunath Sripadarao - 2011-01-14 03:30 Hello, Can you try running the script excluding the Master Node without GPU. For example if node5 is the master node, can you try with # /opt/vizstack/sbin/vs-configure-system -s slurm node[1-4] Regards, - Manju --------------------------------------------------------------------------- On Thu, 2011-01-13 at 16:33 +0000, Cristi Mocan wrote: > Hello, > > After I followed the steps from the admin guide - in order to use > Vizstack in my cluster: ( 1 Master Node -without GPU and 5 clients - > nodes with > GPUs: NVIDIA GTS 8800), install Munge, Slurm, Vizstack. > > Slurm works fine and the connection is good: > > 5 idle node[1-5] > > Next step, I want to configure Vizstack with SLURM: > > I run this command on the master node: > > # /opt/vizstack/sbin/vs-configure-system -s slurm node[1-5] > > > > Result: > > > Processing Node 'node1'... > Errors happened while trying to get the configuration of node 'node1'. > Reason: > > Traceback (most recent call last): > File "/opt/vizstack/sbin/vs-generate-node-config", line 304, in <module> > gpuInfo = parse_gpuInfo(temp_xconfig) > File "/opt/vizstack/sbin/vs-generate-node-config", line 168, in > parse_gpuInfo > busId = "PCI:%d:%02d:%01d"%(int(busIdParts[0], 16), > int(busIdParts[1], 16), int(busIdParts[2], 16)) > IndexError: list index out of range > srun: error: node1: task 0: Exited with exit code 1 > > > Please fix the above errors & run this tool again > > > > I have to mention that I have used nvidia-settings: version 260.19.21 > and the latest Vizstack 1.1-3 released. On all the machines I have the OS: > Ubuntu 10.04. I have used ssh in order to connect to each machine in > my cluster - running in separated terminals. > > I also tried to run: > # /opt/vizstack/sbin/vs-configure-standalone > on one of the client nodes, but I received the same result. > > > Can you help me in order to resolve this problem ? > > Thank you in advance ! ------------------------------------------------------------------------------ Protect Your Site and Customers from Malware Attacks Learn about various malware tactics and how to avoid them. Understand malware threats, the impact they can have on your business, and how you can protect your company and customers by using code signing. http://p.sf.net/sfu/oracle-sfdevnl _______________________________________________ vizstack-users mailing list viz...@li... https://lists.sourceforge.net/lists/listinfo/vizstack-users |
|
From: Cristi M. <Cri...@ne...> - 2011-01-14 11:15:40
|
Hello again,
Unfortunately it doesn't work like this - either with:
# /opt/vizstack/sbin/vs-configure-system -s slurm node[1-4]
or
# /opt/vizstack/sbin/vs-configure-standalone - on the master node(with GPU)
I received the same result:
> Processing Node 'node5'...
> Errors happened while trying to get the configuration of node 'node5'.
> Reason:
>
> Traceback (most recent call last):
> File "/opt/vizstack/sbin/vs-generate-node-config", line 304, in <module>
> gpuInfo = parse_gpuInfo(temp_xconfig)
> File "/opt/vizstack/sbin/vs-generate-node-config", line 168, in
> parse_gpuInfo
> busId = "PCI:%d:%02d:%01d"%(int(busIdParts[0], 16), int(busIdParts[1],
> 16), int(busIdParts[2], 16))
> IndexError: list index out of range
> srun: error: node1: task 0: Exited with exit code 1
>
>
> Please fix the above errors & run this tool again
Do you have other solutions ? I am using vistack_1.1-3_ubuntu9.10.amb64.deb
Thank you in advance.
Cristi
---------------------------------------------------------------------------
From: Manjunath Sripadarao - 2011-01-14 03:30
Hello,
Can you try running the script excluding the Master Node without
GPU. For example if node5 is the master node,
can you try with
# /opt/vizstack/sbin/vs-configure-system -s slurm node[1-4]
Regards,
- Manju
---------------------------------------------------------------------------
On Thu, 2011-01-13 at 16:33 +0000, Cristi Mocan wrote:
> Hello,
>
> After I followed the steps from the admin guide - in order to use Vizstack
> in my cluster: ( 1 Master Node -without GPU and 5 clients - nodes with
> GPUs: NVIDIA GTS 8800), install Munge, Slurm, Vizstack.
>
> Slurm works fine and the connection is good:
>
> 5 idle node[1-5]
>
> Next step, I want to configure Vizstack with SLURM:
>
> I run this command on the master node:
>
> # /opt/vizstack/sbin/vs-configure-system -s slurm node[1-5]
>
>
>
> Result:
>
>
> Processing Node 'node1'...
> Errors happened while trying to get the configuration of node 'node1'.
> Reason:
>
> Traceback (most recent call last):
> File "/opt/vizstack/sbin/vs-generate-node-config", line 304, in <module>
> gpuInfo = parse_gpuInfo(temp_xconfig)
> File "/opt/vizstack/sbin/vs-generate-node-config", line 168, in
> parse_gpuInfo
> busId = "PCI:%d:%02d:%01d"%(int(busIdParts[0], 16), int(busIdParts[1],
> 16), int(busIdParts[2], 16))
> IndexError: list index out of range
> srun: error: node1: task 0: Exited with exit code 1
>
>
> Please fix the above errors & run this tool again
>
>
>
> I have to mention that I have used nvidia-settings: version 260.19.21 and
> the latest Vizstack 1.1-3 released. On all the machines I have the OS:
> Ubuntu 10.04. I have used ssh in order to connect to each machine in my
> cluster - running in separated terminals.
>
> I also tried to run:
> # /opt/vizstack/sbin/vs-configure-standalone
> on one of the client nodes, but I received the same result.
>
>
> Can you help me in order to resolve this problem ?
>
> Thank you in advance !
|
|
From: Manjunath S. <man...@hp...> - 2011-01-14 03:30:27
|
Hello,
Can you try running the script excluding the Master Node without
GPU. For example if node5 is the master node,
can you try with
# /opt/vizstack/sbin/vs-configure-system -s slurm node[1-4]
Regards,
- Manju
On Thu, 2011-01-13 at 16:33 +0000, Cristi Mocan wrote:
> Hello,
>
> After I followed the steps from the admin guide - in order to use Vizstack
> in my cluster: ( 1 Master Node -without GPU and 5 clients - nodes with
> GPUs: NVIDIA GTS 8800), install Munge, Slurm, Vizstack.
>
> Slurm works fine and the connection is good:
>
> 5 idle node[1-5]
>
> Next step, I want to configure Vizstack with SLURM:
>
> I run this command on the master node:
>
> # /opt/vizstack/sbin/vs-configure-system -s slurm node[1-5]
>
>
>
> Result:
>
>
> Processing Node 'node1'...
> Errors happened while trying to get the configuration of node 'node1'.
> Reason:
>
> Traceback (most recent call last):
> File "/opt/vizstack/sbin/vs-generate-node-config", line 304, in <module>
> gpuInfo = parse_gpuInfo(temp_xconfig)
> File "/opt/vizstack/sbin/vs-generate-node-config", line 168, in
> parse_gpuInfo
> busId = "PCI:%d:%02d:%01d"%(int(busIdParts[0], 16), int(busIdParts[1],
> 16), int(busIdParts[2], 16))
> IndexError: list index out of range
> srun: error: node1: task 0: Exited with exit code 1
>
>
> Please fix the above errors & run this tool again
>
>
>
> I have to mention that I have used nvidia-settings: version 260.19.21 and
> the latest Vizstack 1.1-3 released. On all the machines I have the OS:
> Ubuntu 10.04. I have used ssh in order to connect to each machine in my
> cluster - running in separated terminals.
>
> I also tried to run:
> # /opt/vizstack/sbin/vs-configure-standalone
> on one of the client nodes, but I received the same result.
>
>
> Can you help me in order to resolve this problem ?
>
> Thank you in advance !
>
>
>
>
>
> ------------------------------------------------------------------------------
> Protect Your Site and Customers from Malware Attacks
> Learn about various malware tactics and how to avoid them. Understand
> malware threats, the impact they can have on your business, and how you
> can protect your company and customers by using code signing.
> http://p.sf.net/sfu/oracle-sfdevnl
> _______________________________________________
> vizstack-users mailing list
> viz...@li...
> https://lists.sourceforge.net/lists/listinfo/vizstack-users
|
|
From: Cristi M. <Cri...@ne...> - 2011-01-13 16:33:21
|
Hello,
After I followed the steps from the admin guide - in order to use Vizstack
in my cluster: ( 1 Master Node -without GPU and 5 clients - nodes with
GPUs: NVIDIA GTS 8800), install Munge, Slurm, Vizstack.
Slurm works fine and the connection is good:
5 idle node[1-5]
Next step, I want to configure Vizstack with SLURM:
I run this command on the master node:
# /opt/vizstack/sbin/vs-configure-system -s slurm node[1-5]
Result:
Processing Node 'node1'...
Errors happened while trying to get the configuration of node 'node1'.
Reason:
Traceback (most recent call last):
File "/opt/vizstack/sbin/vs-generate-node-config", line 304, in <module>
gpuInfo = parse_gpuInfo(temp_xconfig)
File "/opt/vizstack/sbin/vs-generate-node-config", line 168, in
parse_gpuInfo
busId = "PCI:%d:%02d:%01d"%(int(busIdParts[0], 16), int(busIdParts[1],
16), int(busIdParts[2], 16))
IndexError: list index out of range
srun: error: node1: task 0: Exited with exit code 1
Please fix the above errors & run this tool again
I have to mention that I have used nvidia-settings: version 260.19.21 and
the latest Vizstack 1.1-3 released. On all the machines I have the OS:
Ubuntu 10.04. I have used ssh in order to connect to each machine in my
cluster - running in separated terminals.
I also tried to run:
# /opt/vizstack/sbin/vs-configure-standalone
on one of the client nodes, but I received the same result.
Can you help me in order to resolve this problem ?
Thank you in advance !
|
|
From: Kumar, S. <shr...@hp...> - 2010-11-02 07:18:03
|
Hi Peter,
Thanks for catching this. The fix for this is:
--- python/vsutil.py (revision 350)
+++ python/vsutil.py (working copy)
@@ -552,7 +552,7 @@
nodeIdx = nodeIdx+1
try:
- newNode.setAllocationBias(int(domutil.getValue(domutil.getChildNode(node, "allocationBias"))))
+ newNode.setAllocationBias(int(domutil.getValue(domutil.getChildNode(node, "weight"))))
except:
pass
I have committed this to SVN as version 351. Updating to this should fix your problem. (Changeset 351 also records the node weights in the log file)
That reminds me : I need to put out 1.1-4 soon - many fixes have made it to the code since 1.1-3 !
Regards
-- Shree
-----Original Message-----
From: Pet...@cs... [mailto:Pet...@cs...]
Sent: Monday, November 01, 2010 11:56 AM
To: viz...@li...
Subject: [vizstack-users] Weight values in node_config.xml not working
Hello,
I have four nodes being managed by Vizstack/SLURM (rviz[1-4]) running on Ubuntu 10.04 and would like to have rviz1 allocated jobs last by Vizstack. According to the user guide I should be able to change the weight value for rviz1 to 1 in node_config.xml although this doesn't seem to have any effect ie rviz1 is still allocated first with viz-tvnc/viz-vgl etc scripts. Any ideas?
I've tried 1.1-3 release and the latest source rev 350.
Cheers,
Peter
Peter Tyson
CSIRO IM&T - Advanced Scientific Computing Gate 5 Normanby Road Clayton Vic 3168 Ph +61 3 9545 2021
------------------------------------------------------------------------------
Nokia and AT&T present the 2010 Calling All Innovators-North America contest Create new apps & games for the Nokia N8 for consumers in U.S. and Canada $10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store http://p.sf.net/sfu/nokia-dev2dev _______________________________________________
vizstack-users mailing list
viz...@li...
https://lists.sourceforge.net/lists/listinfo/vizstack-users
|
