BURP not respecting other project slots
Message boards :
Client :
BURP not respecting other project slots
Message board moderation
| Author | Message |
|---|---|
 Contact Send message Joined: 7 Mar 05 Posts: 19 Credit: 35,255 RAC: 3 |
Interesting stuff. <a href="http://pirates.vassar.edu/forum_thread.php?id=378">http://pirates.vassar.edu/forum_thread.php?id=378</a> <a href="http://predictor.scripps.edu/forum_thread.php?id=1631">http://predictor.scripps.edu/forum_thread.php?id=1631</a> |
|
Janus Volunteer moderator Project administrator  Send message Joined: 16 Jun 04 Posts: 4570 Credit: 2,100,463 RAC: 8 |
Thanks for bringing this up. BURP does not, I repeat NOT, work well with other projects. Running BURP (or any other pre-alpha stage project) alongside with other projects is like asking for trouble. Basicly the client ignores any requests to suspend, resume or quit since that part of the client simply hasn't been written yet. How the stderr output can end up with other projects is a mystery to me, but it only emphasizes this further: BURP does not work well with other projects at the moment, period. |
|
RDC Send message Joined: 25 Mar 05 Posts: 18 Credit: 4,117 RAC: 0 |
<blockquote>How the stderr output can end up with other projects is a mystery to me, but it only emphasizes this further: BURP does not work well with other projects at the moment, period.</blockquote> It could relate how corrupted the Windows swap file is and how it is writing the swap file. I have no idea as to how often people reboot their systems but if I let mine run for 4 or 5 days without a reboot, I begin to get some strange things that occur in saving files due to swap file corruption. This is just an observation on my part from what I've experienced but maybe something is in it that can help. I run 4 other projects and haven't had this problem yet as far as BURP goes. I had thought that the suspend and resume issues had been corrected as I had forgot to suspend all other projects when doing BURP on the 127 test and it switched in the middle of a BURP WU to a SETI WU. I suspended SETI at that point and it went back to BURP and no issues on either the BURP WU or the SETI WU and both uploaded fine without errors. BURP did <i>appear</i> to stop and pick up where it left off but I say again, it appeared that way to me. My PC had been recently rebooted a couple of hours prior to this occurring. I have had an Einstein WU corrupt a SETI WU in the past and the PC was up for at least 4 days at that point with no reboot. From what I can see, swap files become corrupted and eventually write dirty data especially if the programs save things in memory and it is pushed out partially and brought back in a few times due to the swap file being full from using the computer for things other than just for BOINC applications. BOINC 4.36 / Win XP Home 2 / 512 memory / AMD Thunderbird 1200mhz |
|
kolch Send message Joined: 23 Apr 05 Posts: 2 Credit: 0 RAC: 0 |
Apparently BURP is leaving some files in the slot after the BURP WU is reported. Later when the slot is used for another project it gets confused and it ends up looking like BURP stole the other projects output. Thats how it reads to me. Not a professional opinion. |
Keck_Komputers Send message Joined: 6 Mar 05 Posts: 94 Credit: 1,384,324 RAC: 7 |
There was a bug in the directory clean up routine. It should be fixed in 4.41. BOINC WIKI   BOINCing since 2002/12/8 |
|
John McLeod VII Send message Joined: 3 Mar 05 Posts: 51 Credit: 79,519 RAC: 0 |
There was a bug in the directory clean up routine. It should be fixed in 4.41. 4.41? If this is really the version number, it is still happening. The problem is that after WU completion, blender.exe is still running in the slot, and therefore cannot be deleted.   BOINC WIKI |
|
Keck_Komputers Send message Joined: 6 Mar 05 Posts: 94 Credit: 1,384,324 RAC: 7 |
There was a bug in the directory clean up routine. It should be fixed in 4.41. Different bug. The one fixed in CCv4.41 allowed directories and files that began with '.' to be deleted. BOINC WIKI BOINCing since 2002/12/8 |
Mr.Pernod Send message Joined: 7 Aug 05 Posts: 33 Credit: 464,805 RAC: 0 |
slot clean-up is still problematic. http://predictor.scripps.edu/result.php?resultid=6484456 after a number of hours blender.exe was still running, even though there was no more work for BURP. due to blender.exe not exiting properly, BOINC is unable to clean out the slot previously occupied by BURP/blender, because the files are still in use.   
|
|
Janus Volunteer moderator Project administrator Send message Joined: 16 Jun 04 Posts: 4570 Credit: 2,100,463 RAC: 8 |
When this happens could you please check the stdout files in the directory, I haven't got any clue to what could cause blender.exe to continue to run outside the BOINC environment. |
|
Mr.Pernod Send message Joined: 7 Aug 05 Posts: 33 Credit: 464,805 RAC: 0 |
When this happens could you please check the stdout files in the directory, I haven't got any clue to what could cause blender.exe to continue to run outside the BOINC environment. I'll zip up the slots next time it happens, just let us know in advance when the next run will be ;) |
|
Mr.Pernod Send message Joined: 7 Aug 05 Posts: 33 Credit: 464,805 RAC: 0 |
blender.exe was still running, over an hour after the last result was reported. again results from another project have been \"processed\" by Blender and reported as computational errors to the other project. Strangely enough, it happened only one of my hosts, not all of them. BURP host 8863 example \"Blended\" Predictor result zip-files: offending slot stdoutdae.txt/stderrdae.txt |
|
Janus Volunteer moderator Project administrator Send message Joined: 16 Jun 04 Posts: 4570 Credit: 2,100,463 RAC: 8 |
Wow! Madness... Thanks for the very comprehensive debug info, this is absolutely crazy. |
|
Rytis Send message Joined: 15 Mar 05 Posts: 77 Credit: 31,788 RAC: 0 |
OK... I accidentally found one such result in PrimeGrid - I was going through the log of erroneous results retuned, and bang!: http://www.primegrid.com/orig/result.php?resultid=340859 PrimeGrid Administrator |
UBT - Halifax--lad Send message Joined: 10 Sep 05 Posts: 74 Credit: 589 RAC: 0 |
Hi Janus, You any idea when you will be able to fix the Blender so that it stops when it is supposed to and lets other projects run there WU\'s correctly?? Join us in Chat (see the forum) Click the Sig  Join UBT |
|
Janus Volunteer moderator Project administrator Send message Joined: 16 Jun 04 Posts: 4570 Credit: 2,100,463 RAC: 8 |
I think I have the issue narrowed down pretty well now so it\'s hopefully \"just\" a matter of recoding the main loop in the client to be much faster and more reponsive. |
|
UBT - Halifax--lad Send message Joined: 10 Sep 05 Posts: 74 Credit: 589 RAC: 0 |
|
|
Mr.Pernod Send message Joined: 7 Aug 05 Posts: 33 Credit: 464,805 RAC: 0 |
I think I have the issue narrowed down pretty well now so it\'s hopefully \"just\" a matter of recoding the main loop in the client to be much faster and more reponsive. any updates? |
|
Janus Volunteer moderator Project administrator Send message Joined: 16 Jun 04 Posts: 4570 Credit: 2,100,463 RAC: 8 |
any updates? Not really, sorry, currently focus is on the mirrors as well as getting the GUI for CATS working. |
|
Mr.Pernod Send message Joined: 7 Aug 05 Posts: 33 Credit: 464,805 RAC: 0 |
any updates? not the update I was hoping for, but an update nonetheless ;) thanks |
|
UBT - Halifax--lad Send message Joined: 10 Sep 05 Posts: 74 Credit: 589 RAC: 0 |
It hardly makes a differance for now as BURP isn\'t really sending out any WU\'s with any luck it will be fixed before it starts sending out WU\'s again Join us in Chat (see the forum) Click the Sig Join UBT |
