Started fold for covid on RPi 3B and I see only 1 task running. It’s having 4 CPU cores then why only 1 task running?
In RPi4 too, it’s only 3 tasks running at a time. It should be 4 ideally.
Any reason for not utilizing all the CPU cores?
Hi there - thanks for helping out with the Fold for Covid effort!
There are two aspects here to consider: memory and the job queue. For devices with 1Gb of memory (3B, some variants of Pi4) we limit the number of tasks running at a time to one, since these tasks are often memory constrained. Also these devices run tasks optimised for smaller amounts of memory, which is a different queue of tasks. Both queues (small and large tasks) rely on Baker Labs adding more tasks to be worked on. It’s possible they are filling the queue. More info here: https://foldforcovid.io/#faqs
Phil
In the early days, we tried to run 4 tasks at a time on both RPi3 and RPi4 (up to 4GB). We experienced reboots, apparent freezes, web interface not accessible - but the RPi apparently still running. So the Balena people experimented with us to reduce the numbers and it was found that for 1GB systems, one task at a time worked reliably. 4GB systems ran reliably with 3 tasks.
The snag was that the 1GB systems were put into their own queue for “small” tasks that they could run. This quite often ran out of work, but if your one is running 1 task then that’s fine. I transferred my 4 x RPi3 and one of my RPi4 to another project which has the work divided up in smaller chunks.
I still have 5 x RPi4 (4 x 4GB + 1 x 8GB) running on here.
I also try be myself forking the repo. I test this options:
- RPi 4 with 2GB
- 3 task
- 2 task
- RPi 4 with 1GB
- 2 task
- RPi 3 with 1GB of course
- 2 task
All these options make the system unstable with many reboots and not finishig task properly.
So, is a low limit but is necessary.
Hi @manuti – thanks for posting your results, and thanks for contributing to FoldForCovid!
All the best,
Hugh