active_task_state: UNINITIALIZED
13 Dec 2019 17:16:54 UTC
Topic 220197
(moderation:
Two work units on my Raspberry Pi 2B are shown as "active_task_state: UNINITIALIZED" and are not being processed, even though one core of the quad-core CPU is available for work. Is there anything I can do to get these work units to execute? I've tried the "update" command, but no change occurs. Thanks for help.
Where do you see that status?
)
Where do you see that status? In Boinc Manager or somewhere else?
What happens if you select all other tasks running or ready to run and suspend them? Only let the 2 tasks with that status be allowed to run.
I'm running this Pi
)
I'm running this Pi headlessly, so I use the keyboard command "boinccmd --get_tasks" to see the status of downloaded work units. "Suspend" is applied to all the work units, and I don't want to stop the three units that are executing properly.
I don't know if the uninitialized state is something that comes from einstein@home or if it's due to something on my set up.
OK, I'm not familiar to
)
OK, I'm not familiar with boinccmd but is this new or might it be that all cached tasks show this state before they are running?
Just guessing but uninitialized sounds like a state before the task is started and would be the state to expect for a task in queue.
If all but the running tasks are suspended then just resume one of the 2 tasks with the uninitialized state and wait for one of the running tasks to complete and then see what happens.
This is not something new;
)
This is not something new; I've seen it before on newly downloaded tasks. Typically, though, four tasks are downloaded and they begin executing immediately, using all four cores of the cpu. Occasionally, the present situation develops, where one of the work units isn't addressed, despite the fact that a cpu-core is available for that task. Eventually, the situation is corrected somehow and all four cores are once again used.
Because I'm accessing the pi remotely through the terminal, I don't know how to start/stop individual tasks.
I think you're correct in your view that the cached tasks are in the "uninitialized" state prior to processing, because once a running task is completed, the formerly uninitialized unit then begins executing. The question I'd like answered is why aren't all four tasks executing immediately if the cpu has the capacity to do so? This flaw is limiting the Pi's usefulness for einstein@home tasks.
So then we've established
)
So then we've established that the state is not the problem but something else is limiting Boinc from running 4 tasks sometimes. The first thing that comes to my mind is settings for RAM-memory use, can't remember the defaults but you might want to review the settings for both when the computer is in use and when it's not and increase the values to see if that changes things.
If that doesn't work then you'll probably have to look into better logging via cc_config.xml and enable more options,<cpu_sched_debug> might help, but read through the documentation in the link and enable the options that seem to fit.
If you have problems interpreting the logs then post them here and I or someone else will help you sort through it.
Many thanks for taking the
)
Many thanks for taking the time to try to help. I'll look at the links you provided.
BTW, I discovered that the boinccmd tool provides a means of accessing tasks remotely (I'd been confining its use to the project as a whole), so I tried to "resume" one of the uninitialized tasks, but it had no effect.
You need to look at your
)
You need to look at your computer preferences and the contents of the global_prefs.xml and global_prefs_override.xml files and make sure you have "use all cpus" set to 100%. Same goes for changing the value of memory to use which is only 50% by default. You need to change that to use up to 90% of memory.
The max_ncpus_pct value was
)
The max_ncpus_pct value was at 100, but changing the memory value to 90% made the difference! All four cores are now busy crunching einstein@home tasks. Thank you so much for your help. Much appreciated!
Einsteinpi wrote:I'm running
)
You can use BOINCtasks on a Win machine (or via Wine if under Linux) to control them. I have a Win laptop that doesn't run BOINC at all but has BT installed and can see the whole farm. I have mixed machines (x64 and Pi's). See https://efmer.com/
Another option is BOINCtui which is a Linux CLI app which can be used in place of BOINC Manager. Its in the Rpi repos.
MarksRpiCluster