overcommitted
29 May 2005 6:56:52 UTC
Topic 189258
(moderation:
my problem is that my client bites off more than it can chew. IOW, it downloads more WUs than it can handle within the deadline. it just finished 1 unit and started the next one, with 2 1/2 days to go until the deadline. the point is, it already downloaded ANOTHER unit with a week to go, but that unit won't finish on time either.
i see 2 options to take care of this problem:
- make sure the client only downloads 1 unit at a time
- or extend the deadline to at least 10 days, better 2 weeks.
overcommitted
)
Are you running different projects?
What is your setting in "connect to network about every xx days"?
What version of Boinc do you use?
What is your CPU time/WU?
Greetings from Belgium
Thierry
Are you running different
)
Are you running different projects?
I run 3 projects: einstein, predictor and seti.
What is your setting in "connect to network about every xx days"?
1
What version of Boinc do you use?
i use boinc menubar 4.43. don't know what version of boinc it is...
What is your CPU time/WU?
different for every project, in the case of einstein it's about 40 hours on a G4-450
Sounds like you have been
)
Sounds like you have been bitten by a startup bug in BOINC, that should be fixed in 4.44. For a brief time at startup the client will not have an estimate of how long the running workunit will take. If it happens to look at how much work you have during this period it will download more because it doesn't think you have enough.
BOINC WIKI
BOINCing since 2002/12/8
Sounds like you have been
)
Sounds like you have been bitten by a startup bug in BOINC, that should be fixed in 4.44. For a brief time at startup the client will not have an estimate of how long the running workunit will take. If it happens to look at how much work you have during this period it will download more because it doesn't think you have enough.
so i should just let it run and it'll sort itself out?
PS. is there a way to remove a WU?
so i should just let it run
)
so i should just let it run and it'll sort itself out?
You can either just let it run and sort it out, or you can upgrade to 4.44, and let it run and sort it out. 4.44 is still in alpha test and carries a higher risk.
PS. is there a way to remove a WU?
Yes, on the work tab in the BOINC Manager there is an abort button. Select the WU to abort and push the abort button.
BOINC WIKI
Yes, on the work tab in the
)
Yes, on the work tab in the BOINC Manager there is an abort button. Select the WU to abort and push the abort button.
BOINC manager? abort button?
PS. i've uploaded a picture, but it doesn't appear in my posts. what's going on?
BOINC manager? abort
)
BOINC manager? abort button?
Go to the "Work" tab, highlight a WU; on version 4.25 or later there will be a button or link on the left to abort the WU.
If you are using 4.19, you R-click the WU and choose Abort from the context menu.
(My memory only goes back a month or so, so I can't recall for sure if there *is* an Abort item in 4.19, but if there, it would be in the context menu. :) )
Ken
you talking about some
)
you talking about some windoze program? i use a mac...
you talking about some
)
you talking about some windoze program? i use a mac...
You may find a BOINC manager for OS X published lately on the download pages. However it has some ... well ... unexpected flaws regarding the interface design and the claimed credits. As I run into an other problem I installed the new program. You have to copy the old files to a new location (/Library/Application Support/Boinc Data) before everything is working again.
As a reward you will get lesser credits for the same amount of CPU time than you got before. Which allready has been less than for other processors before. Just have a look at http://einsteinathome.org/workunit/1148377 .
Besides that the overcomitted- message is still coming up.
so i should just let it run
)
so i should just let it run and it'll sort itself out?
You can either just let it run and sort it out
ok, i just let it run,. hoping it would sort itself out, but it didn't. at one point my client went into panic mode, processing the 2 einstein units first. then, for the next 8 days or so, it kept seti and predictor running, to catch up on their share of the CPU time. now it has reactivated einstein, and downloaded 2 units again. those won't finish on time unless the client goes into panic mode again, and the whole crap will start again, and again. and again...
an einstein unit takes about 48 hours of CPU time on my system, for 2 units that would be 4 days. now consider that einstein gets only 40% of my free CPU time (40% for predictor and 20% for seti), it's not rocket science to see that the deadline (1 week) is too tight.
now the question is, what can i do? it would be ideal if i could restrict it to one unit at a time. is that possible?