Message boards :
Questions and problems :
6.6.38 problems
Message board moderation
Author | Message |
---|---|
Send message Joined: 2 Dec 06 Posts: 69 |
6.6.38 is now the recommended version so I upgraded to it on a couple of machines. Both machines have Intel integrated graphics so no CUDA available. Machine 1: Vista 64 bit, C2Q Q6600, No Cuda GPU After install, it would run for a few seconds (could see the 4 workunit processes running in task manager) and then just go away. Tried different options and finally had to wipe BOINC and install a non-service install. Seems to be working ok now. Machine 2: Vista 32 bit, C2D 6420, 3.x GB ram, no CUDA gpu Upgrade wasn't too painful on this machine. It's running as a service. Left machine alone for a few hours and came back to find that BOINC wasn't running. Downgraded back to 6.6.36 and all is well. Something needs to be done about this. 6.6.36 has work fetch problems which cause the server to refuse to send work, even when it would obviously be able to complete it on time. 6.6.36 also tends to underfetch work, even when the server gives it work if asked for. 6.6.38 has serious problems, especially on 64 bit Vista. There really needs to be a stable, reliable, version of the client. BTW, I've seen some posts in project message boards that seem to indicate that others are having similar problems with 6.6.38 and 6.10.x. Thanks, David |
Send message Joined: 29 Aug 05 Posts: 15480 |
Without error messages from either BOINC itself (to be found in the stderrdae.txt file in your Data directory) or by Windows (to be found in your Event Log), there's not much to go on to try to fix this. Alternatively, download the symbol files (see DebugWinClient) and run BOINC through the Windows debugger Windbg. Then... Right-click My Computer->Properties Advanced Environment variables Add New System variable. Name it _NT_SYMBOL_PATH Give it this value: srv*C:\windows\symbols*http://msdl.microsoft.com/download/symbols;srv*c:\windows\symbols*http://boinc.berkeley.edu/symstore/ Click OK, OK again Install the symbols package if you haven't done so. Just let it unpack to C:\Windows\Symbols Now start Windbg (Start->Programs->Debugging Tools for Windows->Windbg) Go File->Symbol Search Path Click Browse Add your BOINC directory (normally C:\Program Files\BOINC\) Click browse Add your BOINC Data directory (on Vista normally C:\ProgramData\BOINC\) click OK. From the menu bar: File->Open Executable. Browse to your BOINC directory (normally C:\Program Files\BOINC) Click on boinc.exe and click on Open. A command line window will open as well as a Windebug window. Minimize the windows command line window. From the menu bar: Debug->Go Now let the debugger run until BOINC crashes. When BOINC has crashed: In the Windebug window, there's a command bar. At the bottom. Click it and type in 'kb' (no quotes), then hit Enter. A stack trace is being build. It'll say **BUSY** When that's done, go Edit->Write Windows text to file. Save this log as my-debug.txt to your documents or somewhere where you can easily find it. File->Exit, you can save the workspace if you want to (but it isn't necessary). Post the log. |
Send message Joined: 2 Dec 06 Posts: 69 |
OK, It crashed on the Vista 64 Bit machine and I saved the std*.txt files before I did anything to the system. You can find them at http://www.booksnbytes.com/Boinc.6.6.38CrashFiles/ There's also a file called after_crash_revert_6.6.36.txt which is the initial log after I downgraded to 6.6.36... This is significant because it shows a successful upload of the file which 6.6.38 failed to upload immediately prior to crash. The dumps cover 2 crashes. Each crash is preceded by file upload errors in stdoutdae.txt. Apparently something restarted boinc after the first crash and it crashed again within seconds. This is the tail end of stdoutdae.txt just after the crash, beginning with the completion of the Lattice project task that crashes 6.6.38 on upload. 06-Oct-2009 10:49:11 [The Lattice Project] Computation for task 310025290.402308572247175.9_2 finished After installing 6.6.36 over the top of 6.6.38, you can see that it successfully uploaded the files. 10/6/2009 12:21:45 PM Starting BOINC client version 6.6.36 for windows_x86_64 You can find the full stderrdae.txt file on the URL I specified but here's the beginning and end of the data for the second crash. All the modloads are in the full file. BOINC Windows Runtime Debugger Version 6.6.38 skipped middle of file but the full file can be found at the url above. *** Dump of the Process Statistics: *** I hope this helps. From what I'm seeing on the Lattice message boards, others are experiencing the same thing. Again, you can find the FULL files at http://www.booksnbytes.com/Boinc.6.6.38CrashFiles/ --David David Ball |
Send message Joined: 29 Aug 05 Posts: 15480 |
Next test. Can you reproduce the problem with the latest BOINC 6.10.13? If you can, I'll inform the developers. If you cannot, then we can assume it's already fixed in the upcoming client. The 6.6 range is out of development, any big problems with it will not be fixed in that version range. |
Send message Joined: 2 Dec 06 Posts: 69 |
OK, I've loaded 6.10.13 and will see what happens. It might be fixed in the 6.10.x series. I found someone on the lattice boards who was running 6.10.6 and instead of a crash, they got the following error. 04/10/2009 17:04:44 The Lattice Project [error] Error reported by file upload server: nbytes missing or negative The WU on 6.10.6 did validate on Lattice though, even with the 2 permanent upload errors so either the file was zero length or it wasn't actually needed by Lattice. So: 6.6.36 - uploads without error 6.6.38 - crashes BOINC client after "Error reported by file upload server: nbytes missing or negative" 6.10.6 - gets permanent upload error but the WU validates. The main problem I see with this is that 6.6.38 is the default/recommended BOINC client to download and that is the version that crashes. Some people running Lattice are getting frustrated and canceling WU's or dropping Lattice because of this. It will take a couple of days for the current Lattice WU to finish on my machine (estimated 45 CPU hours to go) that I loaded 6.10.13 on. I will post to this thread if it has a problem. -- David David Ball |
Send message Joined: 2 Dec 06 Posts: 69 |
OK, I loaded 6.10.13 on a second machine that had a Lattice WU near completion. It has now completed, uploaded, and validated. Here's the section of the log pertaining to the upload. 10/6/2009 5:36:47 PM The Lattice Project Computation for task 16466160.6849746529756418.1_0 finished It still gets the error from the upload server and it logs a permanent upload error, while 6.6.36 didn't get an error at all. It doesn't crash. It reports the WU and the WU validates with a quorum of 2. It looks like it might still be sending a bad record to the server but it recovers and doesn't crash now. Someone should probably try to figure out why it's getting that error from the server when 6.6.36 didn't. I'm sticking with 6.10.13 on those 2 machines. Thanks, David David Ball |
Send message Joined: 29 Aug 05 Posts: 15480 |
OK, I have kept the developers up-to-date about this thread. The answer is really simple: Lattice should update their server software. The "nbytes missing or negative" is a server error. So why will 6.6.36 upload & report where later versions won't do this, crash on trying or immediately discard of the problematic result files? Because of changes to the client that it doesn't send back garbage results etc. The main gist is the following changes. The code before these changes was giving problems to other projects with 6.6.36 and before. Change Log for 6.6.38 wrote: David 16 July 2009 |
Send message Joined: 18 Nov 08 Posts: 9 |
I am having this problem as well, i have several completed lattice projects, and i dont want to have to delete them. where can i find a previous version of boinc to re-download, or when do you think the problem will be fixed? |
Send message Joined: 20 Dec 07 Posts: 1069 |
|
Send message Joined: 10 Oct 09 Posts: 32 |
I had problems uploading these SZTAKI results with 6.6.38: http://szdg.lpds.sztaki.hu/szdg/result.php?resultid=10415951 http://szdg.lpds.sztaki.hu/szdg/result.php?resultid=10417381 SZTAKI results generate several files for upload - one of which is apparently empty (i.e zero bytes) and I believe the problem occurred when BOINC attempted to upload the empty file. When this happened, the BOINC client disconnected from BOINC manager. All attempts to reconnect failed - including BOINC restarts and BOINC "repair" installations. But, reverting back to 6.6.36 fixed it. Note that I had the same problem with both my laptop and desktop computers. |
Send message Joined: 10 Oct 09 Posts: 32 |
Another instance of the problem with SZTAKI posted here. So I tend to agree with the comment below - 6.6.38 should not be the recommended version. I would also note that the BOINC version history page does not yet acknowledge the existence of 6.6.38. The main problem I see with this is that 6.6.38 is the default/recommended BOINC client to download and that is the version that crashes. Some people running Lattice are getting frustrated and canceling WU's or dropping Lattice because of this. |
Send message Joined: 29 Aug 05 Posts: 15480 |
6.6.36 has a problem where the scheduler will eventually answer that your work request cannot go through, as your BOINC is on for so much percent, with calculation enabled 100%. 6.6.38 has a fix for that on the client side. Just because some people at one project have problems uploading, doesn't mean BOINC has to change its recommended version to something lower with bugs that hit more people at more projects, when the solution for the project with the problems now is simple: update their server software. (*) Seeing SZTAKI their global preferences, they don't even have BOINC 6 preferences support, meaning their server software is most probably still at the 5.11 release that they speak about in this thread. You could just as well revert back to BOINC 5.10.45 then and say that's better off made recommended. (*) I see that the Lattice project has just updated their server software. |
Send message Joined: 10 Oct 09 Posts: 32 |
(*) I see that the Lattice project has just updated their server software. If that means SZTAKI is the lone hold-out, then I agree. But . . . 6.6.36 has a problem where the scheduler will eventually answer that your work request cannot go through, as your BOINC is on for so much percent, with calculation enabled 100%. 6.6.38 has a fix for that on the client side.. . . that was an annoyance. This problem causes 6.6.38 to crash. While that may not be enough for it to rate a "MAY BE UNSTABLE" label, in my book, the "Recommended" rating is questionable. You could just as well revert back to BOINC 5.10.45 then and say that's better off made recommended. When I started with BOINC, I think the recommended version was 4.06. And, I generally upgrade when a new recommended version is released. Having done that, I have seen a lot of improvements to the program and also a lot of minor glitches. 5.10.45 had glitches and so does 6.6.36. Progress isn't always smooth. But I had come to expect a level of stability with the BOINC "recommended version" that 6.6.38 just didn't live up to. That is why I finally joined the BOINC forum and made my report. |
Send message Joined: 29 Aug 05 Posts: 15480 |
. . . that was an annoyance. This problem causes 6.6.38 to crash. While that may not be enough for it to rate a "MAY BE UNSTABLE" label, in my book, the "Recommended" rating is questionable. If you're new to the scene, a BOINC version that doesn't fetch work from your projects due to it being stuck on "calculations enabled 100% of that" is not an annoyance, but something you want to get rid of. As in Add/remove Programs->Uninstall BOINC. Do better next time. This affected all projects out there. For you too, try BOINC 6.10.13. It's the release candidate for the 6.10 series. If it also has crashing problems, let me know and I'll notify the developers. But until that time, there is some responsibility for the Projects as well to keep their server software up-to-date, so try bombarding the admin there with requests to get the 2 year old server software updated. |
Send message Joined: 10 Oct 09 Posts: 32 |
. . . that was an annoyance. This problem causes 6.6.38 to crash. While that may not be enough for it to rate a "MAY BE UNSTABLE" label, in my book, the "Recommended" rating is questionable. As I tried to say earlier, if the 6.6.38 problem is now isolated to SZTAKI, I agree with you. But, having experienced the "calculations enabled 100% of that" problem with 6.6.36, I still think that "annoyance" is an apt descriptor. The point is, neither problem is something that should be in a "recommended" release. (And, in that regard, maybe 5.10.45 was better.) ;-) For you too, try BOINC 6.10.13. It's the release candidate for the 6.10 series. If it also has crashing problems, let me know and I'll notify the developers. I'll do it sometime soon (with SZTAKI) and report back. But until that time, there is some responsibility for the Projects as well to keep their server software up-to-date, so try bombarding the admin there with requests to get the 2 year old server software updated. Knowing what I do about the admins/devs at SZTAKI, I am surprised to hear their software is only 2 years out of date. ;-) But I'll do what I can to prod them and also to warn SZTAKI users against upgrading to 6.6.38. |
Send message Joined: 29 Aug 05 Posts: 15480 |
Knowing what I do about the admins/devs at SZTAKI, I am surprised to hear their software is only 2 years out of date. ;-) But I'll do what I can to prod them and also to warn SZTAKI users against upgrading to 6.6.38. Yes, Adam is a bit slow. But that's fair. :-) |
Send message Joined: 12 Oct 09 Posts: 1 |
I just "upgraded" to 6.6.38 and now my Vista machine won't communicate with the server. Version 6.6.38 works fine on my XT machine. |
Send message Joined: 29 Aug 05 Posts: 15480 |
I just "upgraded" to 6.6.38 and now my Vista machine won't communicate with the server. Which server? What are your messages, if any? |
Send message Joined: 10 Oct 09 Posts: 32 |
For you too, try BOINC 6.10.13. It's the release candidate for the 6.10 series. If it also has crashing problems, let me know and I'll notify the developers. Looks like 6.10.13 works like it should with SZTAKI - at least it did with this result. Here are the messages related to its upload: 10/12/2009 7:04:01 PM SZTAKI Desktop Grid Started upload of caa03215-1b38-489c-abd4-252c594ad6ff_6f4d3cfd-b2ab-4430-b46d-7f9eadb8435f_450036_2_0 |
Send message Joined: 15 Oct 09 Posts: 5 |
I am not sure what is going on but my boinc acount keeps being lost. I will create one then the next time I try to login it says that there is no account. I have 2 computers. both have v...38. one works. My new computer however keeps trying to connect but can't. I have uninstalled & reinstalled multiple times. it will work for a little while then start the whole thing over. When it disconnects I can't access anything. No settings, no buttons. All greyed out. I am realy curious why the program is getting worse not better. It is almost treating me as an invalid user or my account is bad. I was using a USB thumb as a cash with eBooster would that have anything to do with it and if it does it should be with the install requirements as at least a warning or something. Hope whatever is plaging the system gets fixed soon. This is realy anoying. Ive wasted hours trying to solve the issue myself. It looks like it is at least being looked at. I put my sysmptoms out to see if mine are different than the rest. Thanks, John |
Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.