LHC@home

It seems that 'Max # jobs No limit' in my project preferences is limited by the server to 20 tasks a host.

No limit uses the setting from computing, other.

Store at least n days of work & store upto an additional.


Boinc will take the run time of tasks and then try to buffer the amount of work that meet the ndays requirement. To give an example:


Tasks take 4hrs, you have store 1 days of work, with no additional buffer, it would have 4 WU's running if you had quad core, boinc would try to buffer 24WU's.

On a 30 core machine with 10 idle cores I only got 20 tasks. On 2 other machines with lesser cores, I also got 20 tasks,
so that's why I concluded it seems 20 is the maximum to get.
It must be a coincidence and probably again the notorious scheduler issue.

Like VBox and GPU-tasks not exiting or not requested in the server reply.

LHC@home	24 Nov 11:36:38	update requested by user	
LHC@home	24 Nov 11:36:39	Sending scheduler request: Requested by user.	
LHC@home	24 Nov 11:36:39	Requesting new tasks for CPU	
LHC@home	24 Nov 11:36:41	Scheduler request completed: got 0 new tasks	
LHC@home	24 Nov 11:36:41	No tasks sent	
LHC@home	24 Nov 11:36:41	No tasks are available for SixTrack	
LHC@home	24 Nov 11:36:41	No tasks are available for sixtracktest	
LHC@home	24 Nov 11:36:41	Message from server: VirtualBox is not installed	
LHC@home	24 Nov 11:36:41	Tasks for AMD/ATI GPU are available, but your preferences are set to not accept them

When the buffer would be filled (it isn't), I would not get 'Requesting new tasks for CPU', but don't need or project not the highest priority.

Strange, here is my break down:

40thread = 18
44thread = 20
28thread = 20
32thread = 20
56thread = 40
24thread = 17
4thread = 16
4thread = 30
12thread = 46

On the 1st 4 machine, these are mixed with other tasks, so looks like unlimited with 1 CPU allows 20 WU max (1st & 6th would probally be 20 limit too), 5th machine I have set to 2 CPU then this doubles limit to 40. The last 3 machine are dedicated to SixTrack, here the limits seem to be BOINC defined based on setting. The configuration of this project is always a challenge with the settings not working consistently for each sub project, like on ATLAS #CPU actually is number of Jobs.

Different subprojects have different behaviour from each other. My 16 core machine runs sixtrack and Atlas (4 core) currently. Atlas maximum task amount is same as core amount (=4, max # jobs = No limit) sixtrack uses Boinc cache settings (has now 80 tasks waiting).

---

Funny to see how quickly those unsent tasks are sucked out. Just before I went to sleep there was something like 200 000 task available, an hour ago maybe 20 000 and now only 6000+. There must be quite a traffic :)

Project-Server with 32547 Tasks for sixtrack, but
Client say no tasks are available.

Project-Server with 32547 Tasks for sixtrack, but
Client say no tasks are available.

same problem here :-(

Yes, I had that too this morning. I have stopped trying.

the server still shows 32547 unsent tasks, same figure as 12 hours ago.
So obviously, there is some disruption in the download process :-(

Got one task at 6 UTC. Timelimit reached for an other Computer,

https://lhcathome.cern.ch/lhcathome/server_status.php

Looks like we are out of Sixtrack work right now.
I got lucky enough to have enough to run now (2:45am)

yes, we did not have enough in the buffer to cover the entire Christmas break.
one of our users resumed submission at the beginning of January - production of SixTrack results is back to full-steam, we are now filling our buffers.
Happy crunching!
A.

I have seen again on 18/01/2020 in my event log:
No tasks are available for SixTrack

Same happened on 17/01/2020 09:45:58 but. later that day I got 2 SixTracks that ran perfectly

Is this still a problem or is it an issue on my PC because I have run for 11 years of LHC using this machine that always has Windows XP?

Best wishes
Geoff

There are no new tasks queued at the moment. What little tasks are distributed are re-sent, because another volunteer didn't complete them in time. Therefore no specific indications of a problem on your side. Just a slow week for work unit creation.

It seems that SixTrack is a placebo, to pour something from one empty vessel into another. Tasks are generated only for 'Leave me alone'.
Apathetic planet.

Hi Geoff Harmer,
what replied by AuxRx is correct - I will add that the presently active SixTrack users are analysing the results you have produced and deciding the next batch of submission. Thanks for your time and CPUs!
Happy crunching,
A.

any rough idea when Sixtrack tasks will be available again?

I've been getting work consistently (ymmv, there doesn't seem to be enough to build a buffer server side).

Sixtrack come in waves. So we have to wait.