| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Website Support
Thread: Deadline issues, flooded with short deadline wu's |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 11
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
As you can see by my number of posts here I typically just float along on here letting BOINC do it's thing and have for 6 years. I'm unsure how or what happens when new work units are retrieved exactly what takes place but I've often observed I've been fed units that need to be completed in a very short period of time. Meaning I've got wu's already in queue that have deadlines a few days out but when it retrieves some new ones when uploading results it will retrieve some that have a deadline only a few hours or maybe a day off. Well today I happened to check what was going on because the BOINC Monitor on my desktop (Win7 gadget) was overfilled with the limit of wu's it can display. Typically it's doing 12 at a time (on an Intel X980) and the gadget can display progress on up to 20 but it was apparent there were far more than that running because the number of green progress bars was only a few of those shown. Apparently I was bombed with a number of them that needed to be completed quickly and now what I have is 24 wu's that are all going to be (already are) late for the deadline. How is this suddenly happening? BOINC runs all the time on this 24/7/365 and suddenly I got swamped with a bunch of wu's that apparently now I'll not get credit for. I had to abort about a half dozen of them today that it hadn't even started or had done little work on already but I hesitate to abort the others because they are all invested in them with quite a bit of time already. Anyone have any suggestions? Is there something I'm unaware of in settings I should change? This is pretty annoying. Some insight would be great on what took place here.
|
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Welcome to the forums, at last.
Repair jobs generally have 4 days deadline and go to reliable hosts that are known for an under 2 day return time i.e. even if you run with a 2 day cache, you'd still have 2 days to spare. Care to post copy from the My Grid > Result Status page of some of these work units- ttyl --//-- |
||
|
|
kffitzgerald
Senior Cruncher USA Joined: Jan 29, 2011 Post Count: 222 Status: Offline Project Badges: 
|
sometimes I get the same collection of shortly due WU's, what I normally do is suspend any running WU's that have a longer deadline and click on the "no new tasks" until some of the shortly due WU's are finished and then resume the others and release the "no new tasks"
as for why this happens.... I have no clue |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
SekeRob, I could definitely post that information. What method would be best?
----------------------------------------[Edit 1 times, last edit by Former Member at Nov 22, 2011 8:27:51 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
In response to kffitzgerald, usually BOINC would take care of doing that all by itself. In the case here I had already running short deadling wu's and then it added more and pulled them off the already running units that it should have kept going on since they had very close deadlines.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Just follow the navigation path I gave, click on the column headers to sort them on deadline and filter on "In Progress", then pick those you just received with those short deadlines.
Manual managing is not necessary. BOINC prioritizes when it thinks its panic time. As said, WCG repair jobs have a 4 day deadline with a host criteria that these devices to whom they're assigned are known to return work within 2 days. Regular work at WCG as 7 or 10 days deadline. --//-- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
SekeRob when I filter by In Progress it's not showing any of these wu's it is currently working on with deadlines today nor the dozen HCC units it has for tomorrows deadline which it will have no problem meeting (they complete very quickly) unless other wu's for the 23rd are tossed at me later.
----------------------------------------Ok, now I filtered by No Reply and it shows them because they are all overdue. I do notice none of them were given to me in the last day or two, in fact all came from the 12th so they've been in queue for awhile. So that makes this even harder for me to understand. I have never had an issue with late wu's before. [Edit 2 times, last edit by Former Member at Nov 22, 2011 9:09:34 PM] |
||
|
|
krakatuk
Advanced Cruncher Germany Joined: Oct 3, 2008 Post Count: 141 Status: Offline Project Badges: 
|
flibidyflob,
----------------------------------------it's pretty strange what you are describing here... X980 running 24/7/365 wouldn't usually have any problems with deadlines, because it's a great fast CPU. Can you give a bit more information: - What exactly is the deadline of the newly received WUs? Is it less than 4 days? - What are your cache settings? ("connect to internet every..." and "additional cache") I don't believe that you have a big cache on a x980 - it would be a huge load of WUs. But even in this case it wouldn't be an issue for Boinc.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Most improbable, lest the system decided to start fooling with us... as if we've not got anything else to do. So when you filter on HCC and Abort or In Progress and sort them be send or return time, you're not seeing these tasks?
----------------------------------------Of course, if you post as flibidyflob and your client was registered to flobidyflib ** those tasks would be shown under that account. :-) What is your cache setting like? I'm using the WCGDAWS tool by pirogue. This tool nicely lists the "Out" time of all returned results, even those that have been moved to archive (provided tool is run prior to archiving which can be as quick as under 24 hours from validation). But for a few exceptions, most I run go back within 30-40 hours. ** Look in the client project tab for registered member name. --//-- edit: added tool link [Edit 1 times, last edit by Former Member at Nov 22, 2011 9:36:23 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Ok, now I filtered by No Reply and it shows them because they are all overdue. I do notice none of them were given to me in the last day or two, in fact all came from the 12th so they've been in queue for awhile. So that makes this even harder for me to understand. I have never had an issue with late wu's before. So it was in fact 7 days for HCC or 10 days for others when they turned No Reply. Can only think of your client loosing it's tasks and then later refetching them at some point in time. You'd have to dig through the stdoutdae.txt file in the BOINC data directory to reconstruct what happened. Loosing can happen for instance when a client is attached to BOINC Account Manager and WCG not properly ticked as default attached or other mishaps, but the log file would tell. --//-- [Edit 1 times, last edit by Former Member at Nov 22, 2011 9:19:26 PM] |
||
|
|
|