Einstein keeps downloading tasks
21 Sep 2021 21:32:39 UTC
Topic 226067
(moderation:
I don't allow new tasks for Einstein@Home, but it keeps downloading them anyway. Changing this setting back and forth doesn't cancel the problem.
It is really frustrating, because there is no way to stop new tasks from appearing.
The only way to stop it, is to stop the project completely.
That means I can maybe compute some tasks but at some point have to cancel 100 tasks in the future.
How can this be fixxed? It only affects Einstein as far as I know.
If I stop individual tasks,
)
If I stop individual tasks, it downloads new tasks as well. I now have hundreds of Einstein tasks that I don't wanna compute. HELP! >_<
Very strange. Can you post
)
Very strange. Can you post the event log?
21-Sep-2021 23:20:13
)
it looks like that... it even says [Einstein@Home] Not requesting tasks: "no new tasks" requested via Manager but it downloads new ones. >_<
21-Sep-2021 20:19:56
)
this is the complete log after I wanted to get some new Einstein tasks until now ^^
Your issue is you just
)
Your issue is you just suspended those Einstein tasks. You needed to abort them and also set NNT.
As long as the server has a record of sending you tasks in its database, it will keep resending them to you even if have NNT set until it receives a task result or the task times out or errors out from too many errors.
It looks like you reset the project at some time too in the log because of getting the img files again but I don't see the master.xml file in the log list.
I would abort all the tasks on your host, get an ack for the aborted tasks and then reset the project.
Then if you actually want to run the cpu tasks, set a very small cache like 0.01 days and enable tasks again.
Usually I would download a
)
Usually I would download a bunch of GPU tasks and as soon as some tasks pop up I set the project to not accept any new tasks. In this case, tasks would keep downloading..
I let it run over night and it seems like the amount of Einstein tasks is shrinking now with no new ones being downloaded.
Maybe it really was a cache problem, cause I updated it to 6 days. previously I would only use one day. So maybe some hidden queue is involved.....
I will see if the problem is fixxed once all tasks are done
BOINC in general, and the
)
BOINC in general, and the Einstein project in particular, works best if you request work 'little and often'.
This is particularly import where Gravity Wave tasks - the ones with names starting 'h1_...' - are concerned. These use a large amount of input data, but Einstein tries to save internet bandwidth by re-using the same data files again, and again, and again, with slightly different search parameters.
So, Einstein tries to allocate new tasks that match the data files you already have. This matching process involves a lot of database searching by the server when you ask for more work: if you ask for a large amount of work all at once, that search will often take longer than your computer is prepared to wait for the reply.
A 'timeout' will occur, and will be logged in your Event Log. The tasks that were so painstakingly assembled by the server to match your particular data files are lost, because you weren't listening when the reply eventually arrived.
If Einstein simply repeated the same painstaking search the next time you requested work, the same thing would happen - you'd never get any work, because every request would time out.
Instead, the server remembers the list that caused the timeout, and feeds it back to you - slowly, 12 tasks at a time. And it'll keep on doing so, until every task has made it through to your computer.
So, this behaviour is entirely normal, and nothing to be worried about. Just let your computer plod through those six days of tasks that were allocated in the first place, until they've all been processed. You can keep an eye on when things will return to normal, by watching the task list for the computer in question on this website - that will show all the tasks that have been allocated to your machine by the server, including those that were lost.
Yeah, seems like I have no
)
Yeah, seems like I have no other choice. I reset the project, deleted all tasks and now it keeps downloading a bunch of them again while not being asked to get new ones. Seems like I need to be careful with Einstein in the future. I want to spread my contribution to projects evenly and those 6 days of GPU computing will be like months that the CPU projects need to catch up to @_@
I just wished there was a common credit model all boinc projects use. It's really hard to compare projects... some like collatz give like billions of credits, some like nanohub give 1 credit per workunit...