0000665: Win/Linux Network Rendering Issue

Notes
(0001819) psor (reporter) 2009-11-28 08:17 edited on: 2009-11-28 08:24	Should be the same as 0000595 and as Dade wrote related to 0000611, 0000605. nb: What is the right syntax to link an URL when adding a note?

(0001820) MikeJ (reporter) 2009-11-28 10:31	595 was supposedly fixed over a month ago. If the currentl downloads are still doing this then either the download page has not been updated with the fixed software or it has not actually been fixed or this is not the same issue (though it would sure seem to be).

(0001821) psor (reporter) 2009-11-28 11:40	The task was being closed because Dade was pointing out that the bug report was related to other tasks. So I guess it's still not fixed yet. ;o))

(0001822) Dade (developer) 2009-11-28 14:37	Psor, you can link another bug by typing a # followed by the bug number, mantis will do the rest, for instance: 0000595 I have fixed the hung-up (note: not a crash) problem in 0000595. May be this bug report is a duplicate but I'm not sure. MikeJ, first of anything else, you should try one of the weekly builds available from http://www.luxrender.net/forum/viewforum.php?f=30 [^] (builds from 23-11-2009 and 22-11-2009 should be ok) to check if the problem has already been fixed or not.

(0001823) MikeJ (reporter) 2009-11-28 22:32	Dade, I can confirm that the new builds do NOT solve the issue. I just downloaded the November 22 build on the Linux machine and the November 23 build on the Windows machine and the behavior is exactly the same. A couple of the additional pieces of information however: Once you hit the minus button on the LuxRender network tab and LuxRender freezes, the network activity between the master and slave computers continues to be quite high while the slave computer continues to crunch numbers. Then, once the slave computer reports back to the master computer, it gets down to the "Validating SID..." line and stops but doesn't terminate the thread or the terminal window. At that point the "Windows has reported an error..." message box comes up on the master Windows computer and when you hit Don't Send LuxRender crashes to desktop. Hope that helps.

(0001824) Dade (developer) 2009-11-29 06:33	What does tell the line before "Validating SID..." ? Can you copy&paste the log here ? Do you have 2 Linux PC ? Can you try to run the same test with both master/slave under Linux ? Everything works fine for me under Linux, I'm going to try to reproduce your problem with the only Windows PC I have: an eeePC ... it is going to be painful to run Lux on an Atom CPU.

(0001825) Dade (developer) 2009-11-29 06:59	I think to have understood what is the problem. I assume you are are using a very slow network connection and a very short update interval ? What is your update interval ? I think your are experiencing this problem because you are removing a slave while Luxrender is downloading the film from that slave. You can check log tab to see when the download start and finish. You should experience the crash only if you do the operation while there is a film downloading ongoing.

(0001827) Dade (developer) 2009-11-29 09:00	I have commited to mercurial a patch for this problem.

(0001828) MikeJ (reporter) 2009-11-29 12:31	Dade, Here are the last several lines of my LuxConsole terminal window. As I said before, the console keeps going after the '-' key is pressed and then gets to reporting back to the Win32 master computer and this is when LuxRender on the Win32 pc crashes. --------------------- [Lux 2009-Nov-29 12:19:46 INFO : 0] 13:41:29 6239 samples/sec 384.797 samples/pix [Lux 2009-Nov-29 12:19:52 INFO : 0] 13:41:35 6399 samples/sec 384.847 samples/pix [Lux 2009-Nov-29 12:19:58 INFO : 0] 13:41:41 6235 samples/sec 384.897 samples/pix [Lux 2009-Nov-29 12:20:03 INFO : 0] 13:41:47 6730 samples/sec 384.94 samples/pix [Lux 2009-Nov-29 12:20:09 INFO : 0] 13:41:52 6296 samples/sec 384.983 samples/pix [Lux 2009-Nov-29 12:20:15 INFO : 0] 13:41:58 5753 samples/sec 385.028 samples/pix [Lux 2009-Nov-29 12:20:20 INFO : 0] 13:42:04 5867 samples/sec 385.069 samples/pix [Lux 2009-Nov-29 12:20:26 INFO : 0] 13:42:10 5566 samples/sec 385.111 samples/pix [Lux 2009-Nov-29 12:20:32 INFO : 0] 13:42:15 5357 samples/sec 385.148 samples/pix [Lux 2009-Nov-29 12:20:38 INFO : 0] Film transmission done (113300 Kbytes sent) [Lux 2009-Nov-29 12:20:38 INFO : 0] Finished film samples transmission [Lux 2009-Nov-29 12:20:38 INFO : 0] Server processing command: 'ServerDisconnect' (hash: 2500584742) [Lux 2009-Nov-29 12:20:38 INFO : 0] Validating SID: 4675_3167_8001_5216 = 4675_3167_8001_5216 [Lux 2009-Nov-29 12:20:38 INFO : 0] 13:42:21 4785 samples/sec 385.187 samples/pix ------------- Unfortunately, I don't have two linux boxes. Please define 'very slow network connection'. I have them both on wifi, and have good reception to the router. The update interval is set at the defauls 180. I have never changed that. Should I try playing with that? I don't know when the film loads? How do I identify that in the log? I don't know what 'mercurial patch' means. Am I supposed to do something? Thanks for all your help Dade! Mike

(0001830) MikeJ (reporter) 2009-11-29 18:27	Okay, I guess 'committing to Mercurial' means a patch is on the way in the next build. :) I have just confirmed that the November 29 build has fixed this bug. This also means that the bug was only in the GUI because the linux build for this week did not have your patch in it but the Windows build did. This does bring up a question for me though (don't know if this is the best place to ask it though): When I hit the '-' key in network tab, the GUI on the master machine freezes (all rendering stops and LuxRender becomes unresponsive) while the slave machine continues to chew on the rendering -- for up to 10 minutes or more. Once the slave computer finally gets to the next recording session it sends the film and then stops normally and LuxRender on the master machine then resumes activity and rendering, it "un-freezes". Is this expected behavior? Thanks for your excellent work David! Mike

(0001831) psor (reporter) 2009-11-29 18:49	What Mike says is what I did explain in a note at 0000595. It's the same issue. And if it's still there then it will be still bugging people. I haven't had time to test it as I'm moving houses.

(0001833) jeanphi (administrator) 2009-11-30 02:12	Mike's behaviour is exactly the issue I expected when I saw Dade's changes to fix the initial bug. Dade: instead of a lock could we rather have a way to interrupt the sending of the film data so that we don't risk blocking for extensive periods of time? When using locks we should try to always check that the time they are held is bounded to a low duration.

(0001834) Dade (developer) 2009-11-30 03:22	Yes, JeanPhi, that is the plan for 0.7 and most of everything we need to give users feedback (i.e. progress bar) about the downloading of the film too. This stuff is listed in feature requests 0000605 and 0000606. MikeJ, 10 minutes for downloading a film is a long time, I guess it happens because of the wireless connection (i.e. slow connection). Keep in mind a Lux's film can be really HUGE, for instance a 1280x1024 image generates a 26MB film. I suggest you to raise the update interval to 15 minutes or more (it will save you a lot of bandwidth usage and some CPU time too).

(0001839) MikeJ (reporter) 2009-11-30 10:36	I will change the interval time on the network tab by simply adding a zero to the default 180 that is already there. Is there a way to change the default interval time in the LuxRender console? Or will I have to reset this every single time? This current issue of freezing is not really a concern for me. I am one user who is not particularly bugged by it -- as long as LuxRender comes back afterwards such that I can continue working/saving the film, etc..so this is not a real showstopper for me. I'm just happy that LuxRender comes back to me now! Mike

(0001840) Dade (developer) 2009-11-30 14:18	Command line option "-i" can be used to set the update interval. Note: the "-i" option must be used on the master (LuxGUI in your case).

(0001841) MikeJ (reporter) 2009-11-30 17:16	How do I use this command line option in LuxBlend? Mike

(0001843) Dade (developer) 2009-12-01 04:54	There is an option under LuxBlend -> Network -> Update interval tab.

(0001846) MikeJ (reporter) 2009-12-01 12:02	Dade, I see no 'Network' or 'Update Interval' tab in LuxBlend. The only thing I can find is the 'Write' setting under the 'Output' tab in LuxBlend. However, this seems to have no effect whatsoever on the Update Interval seeting in the GUI. No matter what I have this set to in LuxBlend the GUI always reports '180' in the Update Interval window of the GUI. Mike

(0001847) Dade (developer) 2009-12-02 01:13	Under the "System" tab, look for the network option.

(0001851) MikeJ (reporter) 2009-12-02 15:48	How about that? Yet another hidden and completely undocumented item in LuxBlend/Render. Thanks Dade. Mike

Issue History
Date Modified	Username	Field	Change
2009-11-28 00:35	MikeJ	New Issue
2009-11-28 08:17	psor	Note Added: 0001819
2009-11-28 08:18	psor	Note Edited: 0001819
2009-11-28 08:24	psor	Note Edited: 0001819
2009-11-28 10:31	MikeJ	Note Added: 0001820
2009-11-28 11:40	psor	Note Added: 0001821
2009-11-28 14:18	Dade	Status	new => assigned
2009-11-28 14:18	Dade	Assigned To	=> Dade
2009-11-28 14:37	Dade	Note Added: 0001822
2009-11-28 22:32	MikeJ	Note Added: 0001823
2009-11-29 06:33	Dade	Note Added: 0001824
2009-11-29 06:59	Dade	Note Added: 0001825
2009-11-29 09:00	Dade	Note Added: 0001827
2009-11-29 12:31	MikeJ	Note Added: 0001828
2009-11-29 18:27	MikeJ	Note Added: 0001830
2009-11-29 18:49	psor	Note Added: 0001831
2009-11-30 02:12	jeanphi	Note Added: 0001833
2009-11-30 03:22	Dade	Note Added: 0001834
2009-11-30 04:24	Dade	Status	assigned => feedback
2009-11-30 10:36	MikeJ	Note Added: 0001839
2009-11-30 14:18	Dade	Note Added: 0001840
2009-11-30 17:16	MikeJ	Note Added: 0001841
2009-12-01 04:54	Dade	Note Added: 0001843
2009-12-01 12:02	MikeJ	Note Added: 0001846
2009-12-02 01:13	Dade	Note Added: 0001847
2009-12-02 15:48	MikeJ	Note Added: 0001851
2009-12-03 10:07	Dade	Status	feedback => resolved
2009-12-03 10:07	Dade	Fixed in Version	=> 0.6.1
2009-12-03 10:07	Dade	Resolution	open => fixed
2009-12-22 01:49	jeanphi	Status	resolved => closed

Relationships