I am happy to experiment! Let me know what I can do!
Ciao,
Bob
Hi @gratefulfrog - this is going to be a bit complicated, and is definitely experimental. If you’re comfortable, here is a very rough outline of the work:
- Clone the rosetta-at-home repo
- You’ll need to edit the docker-compose.yml file and mark the
boinc-clientservice asprivileged; this page has an example of how to do that - You’ll need to modify boinc-client/start.sh to include instructions to enable zram and configure it. The Arch Linux wiki has instructions on how to do that.
- Finally, you’ll need to build and deploy the application. If you haven’t done this before, there’s a tutorial here.
Like I say, this is definitely experimental, and may not yield any increase in performance for Boinc – changing zram settings may make the device look like it has more RAM, but will also cause a number of performance tradeoffs. If you’re not comfortable with that, or haven’t gone through building an application with Balena before, it would be best to stick with the default foldforcovid image.
All the best,
Hugh
@saintaardvark ok, looks straightforward. I 'll try it tomorrow, I doubt this will make a difference, but it will be fun playing around with it.
The default image is basically useless on an rpi 3B+ since all of mine just sit idle all day long.
If this could get all those Rpi3’s on line a functioning, then we would have made a step forward. and everyone could take advantage.
I’ll let you know.
Maybe we should take this discussion out of the forum thread?
Anyway, we’ll see that tomorrow!
Ciao!
1 Like
its worth checking you have the LATEST code from github the latest version now checks for LOW RAM and only runs LOW RAM jobs
my LOW RAM pi3 (<1G) and pi4 (<2G) are now both running LOW RAM jobs
$ balena ssh 0fb8ff8
=============================================================
Welcome to balenaOS
=============================================================
root@0fb8ff8:~# free
total used free shared buff/cache available
Mem: 990024 584844 60300 38184 344880 522468
Swap: 0 0 0
$ balena ssh 6e02c44
=============================================================
Welcome to balenaOS
=============================================================
root@6e02c44:~# free
total used free shared buff/cache available
Mem: 1953756 915420 239624 37960 798712 1123376
Swap: 0 0 0Hi!
Which release are you running from github? Did you use the installer or build it locally?
I tried to flash:
balena-cloud-rosetta-at-home-arm-raspberrypi3-64-2.47.0+rev1-v10.6.27-1344795.img.zip
but I get the error "invalid or missing account key. To fix remove and add this project.
I have no idea how to fix that…
Thanks for any help!
HI!
I deployed the ZRAM version with 2GB and then with 4GB but that doesn’t seem to change anything.
Full log of my actions is available here
The device logs still show :
01.05.20 14:26:44 (+0200) boinc-client 01-May-2020 12:26:44 [Rosetta@home] Rosetta needs 1716.61 MB RAM but only 918.48 MB is available for use
This is the public URL of my zram version device: https://7b48587c309ccdb391e63002e2323d49.balena-devices.com/
Well that sucks.
I was going to try that myself because I have the same problem and I am running only 1 RPi 3b+ .
I would like to know however how long it might take for it to get a task or if certain times of the day would be better for the servers which send the tasks? I dunno.
At least I’ll run F@H on my AMD A9 when I can (not an amazing cpu) or get someone I know to run it on his beefy RTX1080 :D.
I have left my 4 RPi 3B+ running for 2 days and didn’t get any tasks, ever. So I tried the ZRAM stuff but that did no good.
I will wait for a reply from @saintaardvark then give up.
On the other hand, I have been running the boinc app on my android phone and that is working perfectly…
Well I just got two tasks around 10 minutes ago actually, so I got lucky…
¯_(ツ)_/¯
Does it help that my Pi was connected via Ethernet? I’m not sure.
Also it is just downloading files so I hope it doesn’t just download then show the RAM error like last time.
Ah well…as long as it appears to be doing work I’m happy.
the procedure i have been following is
$ git clone git@github.com:balenalabs/rosetta-at-home.git
$ cd rosetta-at-home
$ git pull
$ balena push MYAPP
on the cloud gui i was setting ACCOUNT_KEY to my boinc ACCOUNT ID
which then auto reboots the container
NOTE OLDER version do not check for LOW RAM
I’ve done the same; but still get the lack of RAM message…
I don’t see a solution; there must simply not be enough low RAM tasks…
I have added 3 x RPi3 to my cluster and two of these are chewing numbers and one is waiting, so there is work out there. All of mine have 16GB SD cards (4 x RPi4 4GB + 3 x RPi3)
At this very moment my 4 RPi 3B+ are all running tasks;
I believe that this is a result of pulling the latest dev version from github.
I tired to create ZRAM using advice posted by @saintaardvark above, but looking with ssh I don’t think it worked. Any more hints about how to configure ZRAM, would be useful.
At least I am now contributing!
Thanks to all that contributed to this!
G
When I try to modprobe zram from the startup.sh, I get this error:
modprobe: ERROR: …/libkmod/libkmod.c:586 kmod_search_moddep() could not open moddep file ‘/lib/modules/4.19.71/modules.dep.bin’
modprobe: FATAL: Module zram not found in directory /lib/modules/4.19.71
but there is no problem when I run the same command from a root shell in the device web interface:
root@7b48587:~# ls /sys/block/ loop0 loop2 loop4 loop6 mmcblk0 ram1 ram11 ram13 ram15 ram3 ram5 ram7 ram9 loop1 loop3 loop5 loop7 ram0 ram10 ram12 ram14 ram2 ram4 ram6 ram8 root@7b48587:~# modprobe zram root@7b48587:~# ls /sys/block/ loop0 loop2 loop4 loop6 mmcblk0 ram1 ram11 ram13 ram15 ram3 ram5 ram7 ram9 loop1 loop3 loop5 loop7 ram0 ram10 ram12 ram14 ram2 ram4 ram6 ram8 zram0 root@7b48587:~#
So I cannot yet say if activating zram is useful or not…
Hi all, just wanted to re-iterate comments from other posts and from the FAQ regarding the Raspberry Pi 3 (all of which have 1gb RAM) and the Pi 4 with 1gb RAM model.
These devices are absolutely capable of running tasks, as I see that some folks have now validated and verified on this thread. They will sit idle until “small” work units are available in the Rosetta@Home work queue, and when there are some available, those Pi’s will grab them and begin crunching.
The issue is that protein folding and the simulations and calculations that are being run are large, complex items, and the majority of them require 2gb of RAM. However, when researchers and scientists at the Baker Lab create the jobs that get submitted to the work queue, some of them do end up being smaller. And, though they are less frequent, when they do get submitted, the Pi3 and 1gb Pi4 can capture them and begin crunching.
The best thing to do is to simply leave the devices running, and when work is available, they will process it.
Hope that helps.
Can confirm, that the Pi 3 got some tasks and it is crunching data. Tried to solve the Zram stuff, but to no avail. Old raspberry is running fine, and will add more to the “fleet”. Thanks balena-team. Keep up the good work 
1 Like
my pi 3b 1gb is working fine, doing tasks daily, still complains about memory but dosnt seem to affect it
1 Like