Advanced search

Forums : Technical Support : Starved of work
Message board moderation

To post messages, you must log in.

AuthorMessage
John Price

Send message
Joined: 2 Mar 08
Posts: 4
Credit: 734,270
RAC: 0
Message 7594 - Posted: 7 Nov 2008, 20:58:21 UTC

I am now getting the message \"Server not sending work as last request too recent\". The problem is I am running a quad core and now one core has no work.
Any work rounds for this?
ID: 7594 · Report as offensive     Reply Quote
sentient_life

Send message
Joined: 6 Jul 07
Posts: 1
Credit: 1,027,643
RAC: 0
Message 7595 - Posted: 7 Nov 2008, 21:23:14 UTC

Enabling new tasks from more than one project is highly recommended.
ID: 7595 · Report as offensive     Reply Quote
Profile Ageless
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 15 Jun 07
Posts: 345
Credit: 50,500
RAC: 0
Message 7596 - Posted: 7 Nov 2008, 21:30:59 UTC - in response to Message 7594.  

I am now getting the message \"Server not sending work as last request too recent\".

You only get that message when you keep pressing Update in BOINC manager, while you\'re being deferred for whatever reason possible. The only workaround is to let BOINC do its job. When it\'s done deferring, it\'ll try again.

Or as said, you can attach to other projects.
Jord.
ID: 7596 · Report as offensive     Reply Quote
Phoneman1

Send message
Joined: 5 Nov 07
Posts: 113
Credit: 3,100,327
RAC: 0
Message 7597 - Posted: 7 Nov 2008, 22:15:34 UTC - in response to Message 7594.  

I am now getting the message \"Server not sending work as last request too recent\". The problem is I am running a quad core and now one core has no work.
Any work rounds for this?


Leave at least 25 seconds before pressing the update button again.

Alternatively, you could try settting \"Maintain enough work for an additional n days\" to a higher number (that is to say, add 1 or 2 to current value - but use with care particularily if you crunch for other projects - it applies to all projects on all your machines.... you could end up with more work than your machines can handle within the deadlines imposed by the projects.

Boinc should top up with work as necessary but you\'ll need to abort any transfers which hang around in a pending state - an on going problem here. Once aborted the next update should get you some more work. If you set Maintain enough for 2 days then you only need to check once a day that there are no downloads pending. Clear those and any others than might hang around in a pending download state when you press update project.

My quads get through about 25 tasks per day each when running Cosmology only, so if you have 50 tasks in your input queue you should not run short work in the next 24 hours even if you get a run of short work units - the Cosmology ones run anything from 2 to 5 hours approximately on my non-over clocked Q6700s.

Phoneman1
ID: 7597 · Report as offensive     Reply Quote
Profile Angus

Send message
Joined: 28 Jul 07
Posts: 71
Credit: 31,250
RAC: 0
Message 7598 - Posted: 8 Nov 2008, 4:00:06 UTC - in response to Message 7597.  
Last modified: 8 Nov 2008, 4:00:23 UTC



Alternatively, you could try settting \"Maintain enough work for an additional n days\" to a higher number (that is to say, add 1 or 2 to current value - but use with care particularily if you crunch for other projects - it applies to all projects on all your machines.... you could end up with more work than your machines can handle within the deadlines imposed by the projects.

Phoneman1


I have never understood this particular attitude \"use with care ...you could end up with more work than your machines can handle\"

If the projects set stupid run times or the scheduler is too stupid to only download the work needed to fill the user\'s desired cache, then too bad. Some of that work will get aborted, either manually or by the projects, and there are lots of people waiting in line to get that resent work.
ID: 7598 · Report as offensive     Reply Quote
Phoneman1

Send message
Joined: 5 Nov 07
Posts: 113
Credit: 3,100,327
RAC: 0
Message 7600 - Posted: 8 Nov 2008, 7:23:00 UTC - in response to Message 7598.  

I have never understood this particular attitude \"use with care ...you could end up with more work than your machines can handle\"


Agreed, if you don\'t report on time then work will be re-issued to some one else.
But Boinc will still crunch the work when it can for you. If the other guy uploads and reports the the result first he\'ll get the credit and you woun\'t get any credit for your late work when you eventually report it.

You will (with this project and I think most) get credit if you report first but late. In this case Boinc will attempt to abort the other guy\'s task if he updates without reporting that task but if his next update includes reporting his copy of the unit then he will also get the credit.

In short reporting late means you run the risk of getting no credit for work done. Some folk don\'t like that hence my warning.

Phoneman1
ID: 7600 · Report as offensive     Reply Quote
sygopet

Send message
Joined: 2 Aug 08
Posts: 27
Credit: 204,771
RAC: 0
Message 7604 - Posted: 8 Nov 2008, 9:26:34 UTC - in response to Message 7600.  

[quote
You will . . . . get credit if you report first but late.
Phoneman1[/quote]

With a quorum of one, as soon as your time is up, the work unit gets errored out and a new one created. You don\'t get any credit if you report late, even by seconds (I know - I have experience!).
One of the (other) problems with Cosmology is the very non-linear \"progress\" with time indicated in Boinc Manager, so that it is difficult to impossible to predict whether a unit will finish processing before the deadline.
ID: 7604 · Report as offensive     Reply Quote
John Price

Send message
Joined: 2 Mar 08
Posts: 4
Credit: 734,270
RAC: 0
Message 7624 - Posted: 11 Nov 2008, 18:50:14 UTC

I\'ve had time to look at my problem a bit more deeply. Firstly I have no problem with single core computers they seem to always be able to get work.
I have two quad core machines that are also running Climateprediction on a 50/50 basis. Climateprediction, for those who know it, has work that can take up to 90 days. It seems that the Boinc manager therefore decides the computer has enough work. If I leave them alone then it completely runs out of cosmology before it asks for more cosmology work. I have set the days work to hold from 0.25 to 1.00 but it has made no difference.
The work round for this is to suspend Climate and then update cosmology which works OK but is a very manual way of running cosmolgy. It still only gives me 4 cosmology units (one for each core,) so not overloaded with work. However, if you need more than one unit using this process, you run in to the issue of \"too soon since last request for work\" it won\'t download more than one and then you have to go thru the problem of waiting, say 60 seconds and then manual update process.

By the way, I still have one out of two downloads hang with 1.88kb of work stalling after downloading 0.36kb. The only solution is cancel this download and update again.

I guess my point is that the idea of BOINC is it is automatic, but with cosmology I am not finding it automatic.
ID: 7624 · Report as offensive     Reply Quote
Profile xPOD

Send message
Joined: 26 Mar 08
Posts: 21
Credit: 267,856,317
RAC: 25,909
Message 7627 - Posted: 12 Nov 2008, 4:56:26 UTC - in response to Message 7624.  

what a bummer, I just brought 112 more CPUs online... and the darn things are ASLEEP! :(
ID: 7627 · Report as offensive     Reply Quote

Forums : Technical Support : Starved of work