1) Forums : Technical Support : Validation errors on camb_boinc2docker (Message 21507)
Posted 6 Sep 2017 by Luigi R.
Post:
I have got a significant number of invalid WUs.

Here are some of these WUs.

https://www.cosmologyathome.org/result.php?resultid=57687968
https://www.cosmologyathome.org/result.php?resultid=57687976
https://www.cosmologyathome.org/result.php?resultid=57687874
2) Forums : News : C@H a BOINC Pentathlon project! (Message 21410)
Posted 6 May 2017 by Luigi R.
Post:
Marius, what is your concept of "soon"? :P

Anyway, are you already tracking results or will they be tracked after you will post the page?
3) Forums : Technical Support : Aborted By Server (Message 21408)
Posted 5 May 2017 by Luigi R.
Post:
As you create a new client, it gets the hostid of another existing client. When and if project server finds out the client is different, all its tasks get cancelled and a new hostid is assigned. You should always start to download your bunker after this differentiation has been done.
4) Forums : Technical Support : Incorrect credit calculation (Message 21042)
Posted 25 Apr 2016 by Luigi R.
Post:
Running 4 tasks, instead of one with 4 cores produces a lot more credits per unit of time.I will let it run for a bit.
If i then switch to 1 task with 4 cores, the credit per unit of time should be very bad and slowly get better, correct?

Yes, it is correct.

Question is: Are credits per unit of time the same for each method when reaching steady state? How log might that be?

I've got a quad core that runs 8 threads at time.

I observed on the boinc2docker app that simulating 64 ncpus in order to run 8 multithreaded tasks, that is 8 virtual threads per 1 cpu thread, produces the same. There are all the problems about CreditNew too.

And I observed on the planck sims app that running 8 threads with the option '--nthreads=1' enabled, that is 1 virtual thread (the application itself not multithreaded) per cpu thread (click & click), produces more credits. I was getting up to +3-4k daily, but after the last update a strange error occurred and I went back to default configuration.
5) Forums : Technical Support : Incorrect credit calculation (Message 21038)
Posted 22 Apr 2016 by Luigi R.
Post:
If I am not wrong, it is caused by CreditNew. It happened to me something like that. Here it is: http://www.cosmologyathome.org/forum_thread.php?id=7363

So there should be some variable that is 'calibrated' in regard to your average runtime. Slowest workunit immediately implies higher credits, but lowering your avg runtime too. So you will get proper credits after your avg will be fixed.
6) Forums : Technical Support : Computational error on planc jobs (Message 21027)
Posted 13 Apr 2016 by Luigi R.
Post:
Ok!

I was using this app_config.xml before
<app_config>
   <app_version>
      <app_name>lsplitsims</app_name>
      <plan_class>vbox64_mt</plan_class>
       <avg_ncpus>1</avg_ncpus>
      <cmdline>--nthreads 1</cmdline>
    </app_version>
</app_config>

because I had observed a better rac ( at least +1-2k cr. <==> +40WUs everyday ). That configuration was good for gpu users too.
It worked till last update.


Now I've no problems running "default" application (multithreading on), but rac is lower.
7) Forums : Technical Support : Computational error on planc jobs (Message 21004)
Posted 8 Apr 2016 by Luigi R.
Post:
Hello, the estimated time of all the workunits on my main client suddenly became 1:58. Now all the workunits started to fail!
8) Forums : Technical Support : Computational error on planc jobs (Message 21003)
Posted 7 Apr 2016 by Luigi R.
Post:
I used some multiple clients to get more work when server get 'dried out', so now I'm emptyimg my work queue.

Main client is running fine and others multiple clients didn't get errors.
I have another multiple client that's behaving like the yesterday's one.

I didn't notice nothing. They are cancelled while running normally and after reaching the time limit.


Some screenshots:
-cancelled task + processes list. [click]
-3 consecutive cancelled task after 59 minutes! So I aborted the remaining tasks. [click]


I had a couple of jobs hitting the elapsed time limit, too. I observed that all those jobs had a very low estimated runtime and were downloaded shortly after the plan_class was changed. Now that the estimated runtime is more realistic, no more jobs are hitting the limit.

Indeed. Estimated runtime is very low on these multiple clients (only 2m53s), maybe because they were running multithreaded workunits some weeks ago. I didn't observed an automatic estimated time's fixing after I reported the first tasks though.
9) Forums : Technical Support : Computational error on planc jobs (Message 20999)
Posted 7 Apr 2016 by Luigi R.
Post:
Same host, other client = no problem.

This is strange.

It appears to be related to boinc software or what?
10) Forums : Technical Support : Computational error on planc jobs (Message 20997)
Posted 6 Apr 2016 by Luigi R.
Post:
Hello, I'm getting a lot of errors while computing with exit status 197 (0xc5) EXIT_TIME_LIMIT_EXCEEDED. Why?

All these tasks get stopped after ~3400s.

Example: http://www.cosmologyathome.org/result.php?resultid=38402245


I'm going to abort the work of this client tonight.
11) Forums : Technical Support : Still getting camb_legacy (Message 20791)
Posted 1 Feb 2016 by Luigi R.
Post:
Why did I get i686 job for a x86_64 machine?

We don't have a 64bit version of camb_legacy for Windows.

I'm on x86_64 GNU/Linux.

Distributor ID:	Ubuntu
Description:	Ubuntu 14.04.3 LTS
Release:	14.04
Codename:	trusty

Linux luis 3.13.0-74-generic #118-Ubuntu SMP Thu Dec 17 22:52:10 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux
12) Forums : General Topics : ncpus simulation and credit system (Message 20790)
Posted 1 Feb 2016 by Luigi R.
Post:
I tried to run all threads for 1 task (as usual for this project).

Here while the task is stuck at 0.100% on BOINC. Total: 45.2% of 800%!


Here while the task is running. Total: 731.6%



I don't know if I discovered warm water, but my host appears to take advantage of running 4 tasks simultaneously.
13) Forums : Technical Support : Still getting camb_legacy (Message 20788)
Posted 1 Feb 2016 by Luigi R.
Post:
Does "Run test applications" help not to fail tasks? Shouldn't test apps be a little unstable? Anyway I updated my preferences.

Why did I get i686 job for a x86_64 machine?
14) Forums : General Topics : ncpus simulation and credit system (Message 20786)
Posted 1 Feb 2016 by Luigi R.
Post:
I noticed that cpu use is lower when a task starts (for 10-20 seconds). It will use 50-75% instead of 200%. So I tried to fill this gap with at least 1 more task. I wanted the remaining % of my cores was used by other tasks.
Total should be 800%.

Here my processes (4 tasks when they are running at all). Total: 781.2%


Here 1 task (the last one) is stuck at 0.100% on BOINC, 3 tasks are running.

There is a little boost on 1st and 2nd tasks. Total: 774.8%

That said, I'm not registering great differences, maybe I've no way to evaluate differences if credits are not reliable. I think I will return to 1 task at time soon or later.
15) Forums : Technical Support : Still getting camb_legacy (Message 20779)
Posted 1 Feb 2016 by Luigi R.
Post:
I have a 64bit OS.
Intel VT-x is enabled, I support VM/LHC projects too.
p_vm_extensions_disabled is set to 0 that means enabled. Should I remove just the same?
16) Forums : General Topics : ncpus simulation and credit system (Message 20778)
Posted 1 Feb 2016 by Luigi R.
Post:
Hello, I'm mainly running boinc2docker app.

I was trying to use more ram and/or getting almost 1 task working while the others are stuck at 0.100% (downloading...).

My cpu is 8-threads... so simulating 16 ncpus would have been fine to run 2 tasks simultaneously. Anyway I started simulating 32 ncpus (4 tasks) and my host is working well, but I noticed my RAC quadruplicated! Is this normal? I would not like to pass for a cheater.



Edit: oh, now (13:00 UTC) I'm getting only twice the ordinary RAC.
17) Forums : Technical Support : Still getting camb_legacy (Message 20769)
Posted 1 Feb 2016 by Luigi R.
Post:
Hello, I've got a 64-bit processor that meets the FAQ requirements.

Yesterday my host run 3 camb_legacy tasks too and it is fine, but maybe it run a wrong architecture task.

My system is x86_64, so why project server fed it on a i686 task? It was invalid after 13h8m25s of elaboration, not fine.
http://www.cosmologyathome.org/result.php?resultid=36235182

x86_64 tasks got validated.
http://www.cosmologyathome.org/result.php?resultid=36231893
http://www.cosmologyathome.org/result.php?resultid=36231945

So I wonder if there is something wrong with work distribution process.