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

N30dG
N30dG
Joined: 29 Feb 16
Posts: 89
Credit: 4,755,917
RAC: 0

Hello, If finished my

Hello,

If finished my testings, and unfortunately I wasn't able to readjust your problem.

  1. flashed a new downloaded image of jessie-lite to my SD-card (from the raspberry website).
  2. apt-get update & upgrade
  3. apt-get install boinc-client
  4. attached boinc to einstein
  5. copy the out-of-place & app_info.xml to projects/einstein.phys.uwm.edu/
  6. systemctl restart boinc-client

Everythink works like expected. Finished some WU. No system-freezes, No restarts.

 

After that I've upgraded to debian stretch. Same result, No freezes, No restarts. Runtimes are equal to jessie. (I've only tested with the out-of-place not the 1.47beta)

 

I really don't know what causes that problem -.- If you using a heatsink + fan without a case I don't think that this problem is heat-related. But as ROBL said: make some temp readings. feeling is not a good measure-method (as weber & fechner discovered Wink)

 

The only thing that I can offer you. If your Internet-connection is good enough and you have some webspace/cloud-storage that is big enough. Make a image of your SD-card, compress it and upload it. Then I can test your setup here on my PI.

Please change passwords (root/user) to somethink that you can give to me and detach boinc from you einstein account, befor making the image.

PG3.14
PG3.14
Joined: 5 Apr 16
Posts: 19
Credit: 2,826,563
RAC: 0

First 4 tasks completed on

First 4 tasks completed on one of my RPI3. Still pending. Times were 20.3k-21k. I will report if they do NOT clear pending.

 

Thanks!!

PorkyPies
PorkyPies
Joined: 27 Apr 16
Posts: 137
Credit: 7,887,573
RAC: 2,882

N30dG wrote:I really don't

N30dG wrote:

I really don't know what causes that problem -.- If you using a heatsink + fan without a case I don't think that this problem is heat-related. But as ROBL said: make some temp readings. feeling is not a good measure-method (as weber & fechner discovered Wink)

vcgencmd says its hovering between 51.5 and 52.6 degrees so I don't think its overheating. It also says its running at 1.2Ghz so certainly doesn't appear to be down-clocking.

 

I have downgraded another Pi to Jessie-Lite and its currently running some 1.47 beta tasks. They look like they will come in around 42k seconds as before. After that i'll put the out-of-place app on it and see if it breaks. My process goes something like this:

1. Write Jessie-Lite (2016-05-27 version) to SD card

2. Boot Pi3 off card, use raspi-config to expand partition, set new password, change device name, set locale and timezone and finally reduce GPU memory to 16Mb. Reboot

3. Do an apt-get update, apt-get upgrade, apt-get remove x11-*, apt-get autoremove

4. Install boinc-client from repo and setup remote_hosts & gui_rpc_auth. Restart BOINC.

5. Using BOINCtasks on windows PC get it attached to Einstein and were off and running.

For the optimised app I ran the cache down, stopped BOINC and then installed it as per instructions. Restarted BOINC and allowed work.

 

N30dG
N30dG
Joined: 29 Feb 16
Posts: 89
Credit: 4,755,917
RAC: 0

@PG3.14: No Problem, I never

@PG3.14:

No Problem, I never saw invalids but feel free to report any issues you've got Smile

 

@PORKYPIES:

Temperatures looks good. I've followed your installation process, with only one exception: I don't waited to ran the cache down, simply abort tasks and switch to out-of-place.  At the moment it run's for over 1h without any issues. But I let it complete at least 8 WU's.

The last think I can imagine to causing that is the powersupply. I've done some measurements about this, using a Agilent 34410A BenchTop-Multimeter (6.5 digit).

Setup:

Multimeter: 10 NLPC; Manual Range 1A; 30min WarmUp-time; 30,000 Values for getting min, max & avg

PI: standart clock&voltage; using your installation-process


Results:

1.47 NEON Beta:

U = 5.08390 V to 5.08394 V

IAVG = 509.53 mA

IMin = 434.1429 mA

IMax = 829.0054 mA


out-of-place:

U = 5.08389 V to 5.08394 V

IAVG = 542.80 mA

IMin = 398.0517 mA

IMax = 999.7005 mA


In this case we should focus on the Imax-Values. The are around 170mA higher as running the 1.47, that have suprised me. I expected not such a big difference. The Values are bouncing between min & max very fast (gone down to 0.006 NLPC to get faster readings). - To lazy to connect my DSO to get some freq-readings. - You should use a good Powersupply that can handle the fast load-ripple, without dropping or increasing the Voltage too much.

I don't know if that's the Problem. I only poking around as long as I can't readjust the Problem. And it seem's that only you have these Problem. But I'm shure that we can find it. It's only a matter of time. Smile


And sorry for my english, It's not good I know...

BQL_FFM
BQL_FFM
Joined: 15 Jul 11
Posts: 4
Credit: 15,715,556
RAC: 0

@N30DG: Excellent job

@N30DG: Excellent job ! 

Thank you.

PorkyPies
PorkyPies
Joined: 27 Apr 16
Posts: 137
Credit: 7,887,573
RAC: 2,882

I suspect the first one I

I suspect the first one I tried it on might be a faulty Pi3. I had some problems with it early on but it seems happy with the project supplied app at the moment. Its #9 in my little cluster.

 

I've just switched #1 over to running the out of place app, its still going after 5 minutes so that's a good sign. Power supply for most of them are the official 2A ones (they recommend 2.5A now). The remaining 3 are plugged into a 40w 5 port charger but none of them have been updated yet. I am running down the caches for the others at the moment and will switch them over as they finish.

N30dG
N30dG
Joined: 29 Feb 16
Posts: 89
Credit: 4,755,917
RAC: 0

Okay, hopefully the new one

Okay, hopefully the new one will do it :)

You maybe can rescue the broken one by give it a little bit more voltage to the core/memory or clocking it down a few Mhz.

robl
robl
Joined: 2 Jan 13
Posts: 1,586
Credit: 1,040,495,747
RAC: 505,091

committed the remaining 3

committed the remaining 3 Pi3s to the new application.

PorkyPies
PorkyPies
Joined: 27 Apr 16
Posts: 137
Credit: 7,887,573
RAC: 2,882

N30dG wrote:Okay, hopefully

N30dG wrote:

Okay, hopefully the new one will do it :)

You maybe can rescue the broken one by give it a little bit more voltage to the core/memory or clocking it down a few Mhz.

#1 has completed its first 4 work units using out of place without any issues. Updated #2 and its still going so looks good. Remainder of cluster running down caches (except the faulty #9). I think I will leave the faulty one running the stock app.

Thanks for the app and your help.

PorkyPies
PorkyPies
Joined: 27 Apr 16
Posts: 137
Credit: 7,887,573
RAC: 2,882

Updated 8 of them. Ended up

Updated 8 of them. Ended up using the in-place after watching a task get bumped out because there wasn't enough memory for the operating system to run stuff.

 

#3 has thrown a couple of signal 11 errors, not sure if that means its a faulty Pi as well.

https://einsteinathome.org/task/576672632

https://einsteinathome.org/task/576542536

 

[edit]

Two minutes later and the Pi has locked up, so looks like another faulty one. Back to running the stock app on it too.

 

Comment viewing options

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