High-Speed Linux-BRP-App for Raspberry Pi 3

KF7IJZ
KF7IJZ
Joined: 27 Feb 15
Posts: 110
Credit: 6,108,311
RAC: 0

Out of 8 Pi 3s, 6 of them

***EDIT****

NEVERMIND - I was attached to https://einsteinathome.org and not the einstein.phys.uwm.edu therefore it never activated the OP

 

Out of 8 Pi 3s, 6 of them started using this updated application immediately.  I have two Pis that no matter what I do (following the installation instructions for the OP version), it WON'T use the downloaded version of the application.  I noticed in the logfiles that it seems to download the "official" BRP 1.47 beta application, but I have NO idea where it is storing it or how the computer is "finding" it to start work units with.  Any thoughts would be appreciated.

 

 

My YouTube Channel: https://www.youtube.com/user/KF7IJZ
Follow me on Twitter: https://twitter.com/KF7IJZ

MarkJ
MarkJ
Joined: 28 Feb 08
Posts: 407
Credit: 66,065,894
RAC: 357

When I ran it I had all sorts

When I ran it I had all sorts of problems. The main one being that the Pi wouldn't have enough memory to run 4 at a time. I would also experience random lock ups and I got a lot of invalids. The app may have improved since then but I would be cautious as with any beta-app.

MarkJ
MarkJ
Joined: 28 Feb 08
Posts: 407
Credit: 66,065,894
RAC: 357

KF7IJZ wrote:Out of 8 Pi 3s,

KF7IJZ wrote:
Out of 8 Pi 3s, 6 of them started using this updated application immediately.  I have two Pis that no matter what I do (following the installation instructions for the OP version), it WON'T use the downloaded version of the application.  I noticed in the logfiles that it seems to download the "official" BRP 1.47 beta application, but I have NO idea where it is storing it or how the computer is "finding" it to start work units with.  Any thoughts would be appreciated.

They should be in /var/lib/boinc-client/projects/einstein.phys.uwm.edu

Stop the boinc-client  (sudo service boinc-client stop)

Make boinc the owner of the contents of the folder (sudo chown boinc:boinc /var/lib/boinc-client/projects/einstein.phys.uwm.edu/*)

Mark the program as executable (sudo chown +x name of executable goes here)

Start up boinc again (sudo service boinc-client start)

MarkJ
MarkJ
Joined: 28 Feb 08
Posts: 407
Credit: 66,065,894
RAC: 357

MarkJMark the program as

Mark the program as executable (sudo chown +x name of executable goes here)

Should say: sudo chmod +x name of executable goes here

KF7IJZ
KF7IJZ
Joined: 27 Feb 15
Posts: 110
Credit: 6,108,311
RAC: 0

Thanks MARKJ - my issue was

Thanks MARKJ - my issue was that I had attached to the wrong URL - when I attached to the long URL, everything worked a treat.  I followed N30dG's instructions and it worked as expected.

I have not had any problems running the OP versions of the app, but all my Pis are running Raspbian Lite and I have set the GPU Memory to 16Mb, so I've had no issues with running 4 clients nor with errors on results.  I also have active cooling and heatsinks on all my Pis.

My YouTube Channel: https://www.youtube.com/user/KF7IJZ
Follow me on Twitter: https://twitter.com/KF7IJZ

Elvis2000
Elvis2000
Joined: 18 May 17
Posts: 5
Credit: 289,313
RAC: 0

This is my first post and I

This is my first post and I apologize up front for its length. I'm also new to the Rpi and Linux world (I'm a long-time Windows user), so my use of proper terminology and understanding of Linux might be a bit lacking.

I got started with BOINC on May 18th. I initially installed the BOINC Manager package using "Add / Remove Software" on a Raspberry Pi 3 (rpi3-2) running Raspbian Jessie with Pixel (April 2017). I eventually worked my way into setting up an Einstein@Home account via BAM! This Rpi has been running since then with no problems; however, since it's using the GUI BOINC package, it's the slowest wih a turnaround time of around 2 days. I intend to move this to Jessie Lite and boinc-client soon.

The next day (May 19), I set up another RPi 3 (rpi3-3) using Jessie Lite and boinc-client from the command line. Using instructions from the end of KF7IJZ's "Raspberry Pi 3 Super Computing Cluster Part 2 - Software Config" video, I attached to Einstein@Home directly rather than using BAM! For the first day or so, everything seemed to run okay; although, I did get a number of invalid results. I accumulated about 250 credits before the real problems set in with this Rpi. I'll address that in more detail below.

A day later (May 20), I set up a third Rpi 3 (rpi3-4) similarly to rpi3-3 (Jessie Lite, boinc-client), except that I followed the instructions at the beginning of this thread and installed "einstein_32bit_0.13-OP.tar". I attached to einstein.phys.uwm.edu and have been running since the beginning with no problems. Despite starting 2 days later, rpi3-4 has accumulated almost 70% more credit than rpi3-2 (1,668 vs 1,000, as of this post) with an average turnaround time of .51 days.

It should be noted that all three of the RPis have been modified to boot and run directly from a USB thumb drive rather than a micro SDcard (using the instructions at raspberrypi.org for booting from a MSD).

So, rpi3-2 is running full Jessie and GUI BOINC Manager with no problems (it's just slow) and rpi3-4 is running Jessie Lite and boinc-client (modified) with no problems (and much faster than rpi3-2).

The other Rpi 3, rpi3-3, has been nothing but problems. I originally had it set up using the "stock" boinc-client installed from the command line and after a day or two it started to freeze up. I had SSH windows open with HTOP and a temperature watch running and I'd come back to find them timed out and the Rpi frozen. Checking the task details on the site, I found lots of computing errors and invalid results. I accumulated 250 credits the first day or so, but nothing more since.

So, I re-imaged the USB stick and installed the modified boinc-client version (like on rpi3-4), but it had the same issues; computing errors and eventual freezing. Fearing the Rpi might be "bad," I installed Stress and ran three 10-minute stress tests. All 4 CPU cores were pegged at 100% and I monitored temps and CPU freqs. It passed the stress test each time. The CPU never throttled (it always stayed at 1200 Mhz) and temps never exceeded 65 degrees C.

As an aside, all of my Rpis have heatsinks and I have a 120mm AC Infinity fan blowing air across the heatsinks; so, under "normal" BOINC loads, the CPU temps rarely exceed 60 degrees C and tend to hover between 55 and 60 degrees.

At one point yesterday, rpi3-3 errored out with the following appearing in the SSH window (it was frozen at this point, but I was able to copy this much to my clipboard):

Message from syslogd@rpi3-3 at May 23 21:14:18 ...0.4 0:00.10 /usr/sbin/ntpd -p
kernel:[ 710.814868] ff60: 00000000 00000000 bc94e000 bcf24800 bcf24800 80c0312c 00000008 80c03198 0 11552 3220 2600 S 0.0 0.3 0:00.21 sshd: pi@pts/0
878 boinc 39 19 204M 201M 4452 S 0.0 20.7 0:00.09 ../../projects/ei
Message from syslogd@rpi3-3 at May 23 21:14:18 ...0.7 0:00.09 ../../projects/ei
kernel:[ 710.824620] ff80: 80c15e86 80c15e86 bc94ffc4 bc94ff98 801988b0 8018caf4 bc94ffb4 bc94ffa8 0 10236 2668 2416 S 0.0 0.3 0:00.38 /lib/systemd/syst
629 root 20 0 2564 1696 1448 S 0.0 0.2 0:00.02 /sbin/dhcpcd -q -
Message from syslogd@rpi3-3 at May 23 21:14:18 ...0.3 0:00.07 avahi-daemon: run
kernel:[ 710.834393] ffa0: 805d8e54 805d73c4 bc94e000 80c0312c 00000008 80c03198 bc94ffdc bc94ffc8

Message from syslogd@rpi3-3 at May 23 21:14:18 ...
kernel:[ 710.844157] ffc0: 80161c60 80198858 00000000 80c7a338 bc94fff4 bc94ffe0 8010dcf0 80161b64

Message from syslogd@rpi3-3 at May 23 21:14:18 ...
kernel:[ 710.853992] ffe0: 3c8e806a 00000055 00000000 bc94fff8 0010196c 8010dbcc 55555555 55555555

Message from syslogd@rpi3-3 at May 23 21:14:18 ...
kernel:[ 710.863998] Code: bad PC value

This means nothing to me, but maybe someone else might see something useful.

On a lark, I decided to switch from booting to a USB thumb drive, so I set up a micro SDcard with the same structure as the USB drive (Jessie Lite, modified boinc-client). It has been running for 18 hours without freezing (with the USB drive, it froze within an hour), but I still got some invalid results. I just checked and I finally got some new credit from rpi3-3 from this latest run, so for now it seems to be functioning properly.

So, to summarize:

- rpi3-2 and rpi3-4 are booting to and running off of USB thumb drives with no problems
- rpi3-2 is using Jessie with Pixel and the GUI BOINC Manager; turnaround time is about 2 days
- rpi3-4 is running Jessie Lite and the modified boinc-client; turnaround time is about .5 days
- rpi3-3 is currently running Jessie Lite and the modified boinc-client from a micro SDcard (rather than a USB thumb drive); it currently shows an average turnaround time of .68 days

Something else I noticed from my computer summary page that's probably relevant: both rpi3-2 and rpi3-4 are running Linux 4.4.50-v7+, but rpi3-3 (the "problem child") is running Linux 4.9.24-v7+. So two of the Pis (running the same Linux kernel) are running off of USB thumb drives with no problem (and one is using the "modified" boinc-client), and the other Pi using the newer kernel had all sorts of problems running off of a USB drive, but so far seems to be working okay running off of a micro SDcard.

Whew! I think that about describes it. Thoughts? Comments? Any ideas as to why rpi3-3 seems to have (had) issues and the others don't?

mmonnin
mmonnin
Joined: 29 May 16
Posts: 265
Credit: 918,466,321
RAC: 580,006

I only have experience on a

I only have experience on a single RPi2. It runs about 22.5 hours per task, around 81k seconds. Altho the last set are around 84k seconds. I don't really look at the errors but of the ones still in the history are all validated. This is at stock clocks and the default BOINC installation and RPi application from E@H.

 

Running off an SD Card or USB stick can be hard on it as they aren't really designed for the constant read/writes of an OS. I have heard people had their cards fail on them while crunching

PorkyPies
PorkyPies
Joined: 27 Apr 16
Posts: 137
Credit: 7,887,448
RAC: 2,885

I had noticed that since the

I had noticed that since the 4.9 kernel came out the frequency of disk writes has reduced. You should see the green led flashing a lot less. Good for microSD cards. I haven't tried the optimised app since the 4.9 kernel came out, but was getting lots of errors the last time I tried it.

If you use bam then you need to attach or detach projects using it rather than directly attaching to a project.

If you have a windows pc you can run BOINCtasks to monitor your bramble as they call them in the Rpi world. You can see all of them in one go.

Elvis2000
Elvis2000
Joined: 18 May 17
Posts: 5
Credit: 289,313
RAC: 0

Well, after running for about

Well, after running for about a day and a half I just had another kernel panic this morning on rpi3-3:


Message from syslogd@rpi3-3 at May 25 10:36:09 ...
kernel:[132927.419160] 5f60: bc155f8c bc155f70 8013aaac 80271eb4 bc154000 bc154010 80108244 bc155fb0

Message from syslogd@rpi3-3 at May 25 10:36:09 ...
kernel:[132927.420959] 5f80: bc155fac bc155f90 8010b838 8013a9ec 00000004 7eb33f34 01c3fb4f 00000006

Message from syslogd@rpi3-3 at May 25 10:36:09 ...
kernel:[132927.422835] 5fa0: 00000000 bc155fb0 801080e8 8010b778 00000000 0002b0d3 00000755 000000ab

Message from syslogd@rpi3-3 at May 25 10:36:09 ...
kernel:[132927.424807] 5fc0: 00000004 7eb33f34 01c3fb4f 00000006 01c47680 00000000 01c148a0 01c3f658

Message from syslogd@rpi3-3 at May 25 10:36:09 ...
kernel:[132927.426875] 5fe0: 00000000 7eb33ccc 000175cc 76de801c 60000010 00000004 00000000 00000000

Message from syslogd@rpi3-3 at May 25 10:36:09 ...
kernel:[132927.438004] Code: e3405008 eb1234ef e5943000 e1a00006 (e3130008)
Segmentation fault

I'm going to try one other thing and if it doesn't work I'll probably just retire this Pi from BOINC. I re-imaged the USB thumb drive and reinstalled the stock boinc-client, but I didn't do an update/upgrade; so, rpi3-3 is now on Linux 4.4.50-v7+ (just like the other two problem-free Pis) rather than 4.9.24-v7+, and it's using the stock boinc-client rather than the optimized app. The other two Pis (in fact, all 5 of my other Pis) run just fine from a USB thumb drive, so I'm going back to that instead of the micro SDcard.

We'll see what happens. So far, it's been running for almost 3 hours without a kernel panic. If it craps out again, I guess I'll have to start treating it as a hardware issue, rather than a software problem.

 

Elvis2000
Elvis2000
Joined: 18 May 17
Posts: 5
Credit: 289,313
RAC: 0

Since my last post, I've had

Since my last post, I've had 2 more kernel panics, but it also managed to accumulate some more credit throughout all of this. I was at 250 on rpi3-3 when the first problems started and I'm at 813 as of the latest crash.

As I mentioned above, I'm going to retire this device from BOINC after I run out the remaining tasks and see if it runs "properly" doing other things. Since I wasn't running the modified app this time, it seems the problems aren't specifically related to that. Maybe it's a different part of the BOINC client; maybe it's totally unrelated to BOINC.  I have two other Pis running BOINC with no problems whatsoever (one of which is running the optimized app); neither has produced a single error from the start -- which points to the issue being this specific device.

Unless I find it's something related to the topic of this post, I guess there's not much else relevant to say. For a $35 part, economics says to just chuck it and move on -- probably not worth the time to fiddle with it. But, I'll probably fiddle with it, anyway.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.