Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 49
|
![]() |
Author |
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18665 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I can't believe this is a bug so what could explain this? I was running 5.4.11 (Windows) and upgraded to 5.8.8. What I saw was WUs getting crunched but I had yet to get any new work. I got down to the one WU being crunched being the only WU in my queue. Still no new work. Using the update button would not get any new work. Even with the one WU suspended - nothing crunching and no deferred communications - the update button would not get any new work. Okay, I know changes were made to the work fetch and scheduler but it looks like my "Connect to server" setting is being ignored. So, just how is this all supposed to work now in 5.8.8? If I let this last WU crunch under 5.8.8, it should finish around midnight. I would rather not take a chance on having my machine sit idle thru the night so I'm going back to 5.4.11 in the meantime.
---------------------------------------- |
||
|
winterM
Cruncher Joined: Nov 16, 2004 Post Count: 42 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I have the same problem.
|
||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Lots of folk are scratching the back of their heads on this, but running 5.8.8 on a C2D i know from 1 week watching 10 in 10 out, that it will get work as late as after finishing the last one or when one of the 2 cores idles. FAAH real lengths are so wild.... last batch varying from under 4 hours to over 8 hours, that the RDCF gets blown off it's socks any time a 8 hour version comes by, when 5 hours were predicted....repeat that a few times and...well u get the picture.
----------------------------------------Exclude FAAH from the all mix batch and after a few days normality resumes. No guarantees, but that's what's happening here. ciao PS: Here's a thread at BOINC Dev to get the feel. http://boinc.berkeley.edu/dev/forum_thread.php?id=1527
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18665 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Sek, just checked. My current DCF is 1.6. Yes I crunch FAAH but I wan't noticing a wide difference between TCs and actual. In fact, I had been noticing that the TC's seemed to be quite accurate lately. Still, it appears to me that my "Connect to server" setting is being completely ignored. I find that hard to believe as that would be a pretty basic flaw.
----------------------------------------The one WU I had was FAAH but report deadline is 2/10 - four days from now. With that as the only WU, how could BOINC think it wouldn't finish a WU with a 9 hour TC in 4 days? I read the thread you referenced and my first impression is that the results we'll get with a "Connect to server" setting different from the default will be too complicated for this to remain a useful option. Is the fact that WCG is a multi-project project a complication here? I would think if I have a "Connect to server" setting of 3.5 and only one WU that is currently crunching with a TC of 9.x hours and a report deadline FOUR days hence, I would get more work hitting the update button. This is going to have to make a LOT more sense before I'll go back to 5.8.8. |
||
|
Dingo
Cruncher Australia Joined: Nov 2, 2005 Post Count: 30 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I was just about to upgrade to 5.8.8 but will give it a bit of time and see if there are any other problems.
---------------------------------------- |
||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Sek, just checked. My current DCF is 1.6. Yes I crunch FAAH but I wan't noticing a wide difference between TCs and actual. ....... This is going to have to make a LOT more sense before I'll go back to 5.8.8. Well this moment the RDCF on the C2D is 1.14 and the connect is 2.5 days, and the preference option is 'always on' and have 3 FAAH's left. The P4 has gone wonky just now.... series of HDC+GC at 1:30 to 2 hours.... hello, 2x 4:30 in between.... no more trickle as the predicted time for the rest in the buffer shoots up with it and a RDCF of 2.7. The GC's remaining have adopted the HDC times now of 4:10 when knowing they will be done in 1:30. Does is make sense? It's just that the WCG multi project, multi lengths within each of projects wrecks havoc on the new system. The bloodied towel is in the ring.... the subject is bedded.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18665 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Sek, just checked. My current DCF is 1.6. Yes I crunch FAAH but I wan't noticing a wide difference between TCs and actual. ....... This is going to have to make a LOT more sense before I'll go back to 5.8.8. Well this moment the RDCF on the C2D is 1.14 and the connect is 2.5 days, and the preference option is 'always on' and have 3 FAAH's left. The P4 has gone wonky just now.... series of HDC+GC at 1:30 to 2 hours.... hello, 2x 4:30 in between.... no more trickle as the predicted time for the rest in the buffer shoots up with it and a RDCF of 2.7. The GC's remaining have adopted the HDC times now of 4:10 when knowing they will be done in 1:30. Does is make sense? It's just that the WCG multi project, multi lengths within each of projects wrecks havoc on the new system. The bloodied towel is in the ring.... the subject is bedded. Sek, clearly there is a method to how this is coded in 5.8.8. However, at this juncture, it does not seem highly intuitive to me. Running multiple projects (in the BOINC sense) can require some coordination and planning by the user in order to avoid unexpected consequences. However, I only crunch WCG. Within WCG, I crunch FAAH and FCG. I have found that crunching HDC appears to introduce enough oddities that I have stopped crunching on it for now. My TC's for FAAH ran between 9-11 hours and FCG was typically 2.5 hours. For the WU's I have been crunching, those have been rather accurate times compared to actual. Usually, when the TCs get thrown off, it tends to be dramatic enough that it takes several days to settle back down. I saw no noticable change in my TC's. Despite all that, the one question I can't find a good answer for is why, with only ONE WU on my machine - the FAAH that was actively crunching - that isn't due for four days and is expected to take only 9-10 hours to complete - why did BOINC not send me any more work and why the Update button would neither. Maybe that's an overly simple view of this but is it unreasonable to wonder why when you have 1/3 of a day's work left. four days left to complete it, and a Connect to Server of 3.5 day, my machine still can't get any new work at all, not even ONE WU? All I want to do is ensure my machine keeps crunching and has adequate work on hand to weather typical outages. |
||
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18665 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Hmmmm.....may be on to something possibly? Reread the thread in the link in Sek's post yet again. Went browsing around in those fora and found this: BOINC core client : RDCF is still going up which seemed to put a "face" on an odd feeling I had. I don't regularly go check on my DCF value but when I have under normal conditions, it always seems to be between 1.0 and 1.3 and yet I'm seeing 1.6 at the moment. The thread linked appears to have this as something fixed in 5.8.9 but that's not "gold" yet. Maybe we just need to wait for that.
----------------------------------------![]() |
||
|
watzkej
Cruncher Joined: Mar 12, 2005 Post Count: 12 Status: Offline |
Hmmmm.....may be on to something possibly? Reread the thread in the link in Sek's post yet again. Went browsing around in those fora and found this: BOINC core client : RDCF is still going up which seemed to put a "face" on an odd feeling I had. I don't regularly go check on my DCF value but when I have under normal conditions, it always seems to be between 1.0 and 1.3 and yet I'm seeing 1.6 at the moment. The thread linked appears to have this as something fixed in 5.8.9 but that's not "gold" yet. Maybe we just need to wait for that. ![]() I would say that if you're really having problems keeping work in your queue and the client isn't just waiting until the last possible moment to go get work then you could try 5.8.9. I've been working with the 5.8.x series for some time and 5.8.9 is actually quite stable. The changes between 5.8.8 and 5.8.9 aren't too major and the Windows testing so far looks good. I wouldn't worry about a massive crash or anything at this point since 5.8.x has been poked and prodded fairly well. You can always reinstall 5.8.8 if there was something that you didn't like. Besides, if 5.8.9 really helps your situation, the feedback posted here could be shot over to the boinc_alpha list (that I work with) so that the devs know that the fixes are indeed going well in the DCF area. - John Watzke |
||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Upgrading the C2D from 5.8.8 to 5.8.9 and see how that goes for the next few days on a monotone diet of FAAH. The post by Jord (Mod of BOINC dev) gives some expectation:
----------------------------------------Message 8045 - Posted 7 Feb 2007 1:12:10 UTC I've now been running 5.8.9 for a couple of hours and see the results already. Primegrid's ETC's are now around 5m 28s and still going down, instead of 8m 40s and going up. :-) ____________ Jord If the 'minor' tweak has that effect, it's 'major' to most.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
![]() |