HELP PLEASE Project NOT obeying preferences!

Message boards : Questions and problems : HELP PLEASE Project NOT obeying preferences!
Message board moderation

To post messages, you must log in.

AuthorMessage
Robster

Send message
Joined: 4 Jul 11
Posts: 21
United Kingdom
Message 38776 - Posted: 4 Jul 2011, 9:13:05 UTC

Hi

I am currently running 3 projects on BOINC.

On one of my machines the climateprediction project takes 80% CPU and heats up my server significantly (80c plus).

Seti and rosetta when they are running all seem to operate on or around the 30% I have set up and the machine runs at 51c much more reasonable.

This is on a Mac Mini with 10.6.8.

ANY help would be appreciated!

Robin
ID: 38776 · Report as offensive
muddocktor

Send message
Joined: 4 Jul 11
Posts: 5
United States
Message 38784 - Posted: 5 Jul 2011, 6:35:27 UTC

Download and install smcFanControl on your Mini. I have been running that on my old Mac Mini since shortly after getting it and it let's me control my fan speed. I have the fan speed jacked up to 100% and my Rosetta crunching temps run around 50-55 C. Apple didn't make fan rampup profile very aggressive on my Mini and before I installed smcFanControl, my temps averaged in the 80's. BTW, I recently upgraded to a T7600 from my old T5600 and I haven't seen any increase in average temps either.

BTW, the original install was done on Leopard and I just upgraded to Snow Leopard and had no problems with the program running correctly.
ID: 38784 · Report as offensive
Robster

Send message
Joined: 4 Jul 11
Posts: 21
United Kingdom
Message 38871 - Posted: 10 Jul 2011, 8:54:24 UTC - in response to Message 38782.  

Hi

Yes, you are spot on.

If I have told BOINC only to allow 30% CPU utilisation I do not expect any project to take 80%.

Having said that I noticed climateprediction errored with a checksum error I think and has never downloaded a unit since, so the problem has not replicated itself.

Robin
ID: 38871 · Report as offensive
Robster

Send message
Joined: 4 Jul 11
Posts: 21
United Kingdom
Message 38872 - Posted: 10 Jul 2011, 8:56:18 UTC - in response to Message 38784.  

Hi

Thanks for the idea but I don't want to run the fans high, I want to keep BOINC CPU utilisation at a level that contributes to the effort but without generating excess heat and noise from my systems.

Make sense?

Robin
ID: 38872 · Report as offensive
Robster

Send message
Joined: 4 Jul 11
Posts: 21
United Kingdom
Message 38885 - Posted: 11 Jul 2011, 17:16:38 UTC - in response to Message 38875.  

Okay, so maybe I misread.

So to be clear, or should I say, clearer!

As far as Resource Share is concerned it is split evenly between all three projects and rotates every 60 minutes.

This works fine.

CPU however, is set to 'use at most 25% of multiprocessors'

and 'use at most 30% of CPU time'

By the may, this is on a Mac Mini with a 2.4 Ghz Core 2 Duo CPU.

Robin
ID: 38885 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 38886 - Posted: 11 Jul 2011, 17:34:38 UTC - in response to Message 38885.  
Last modified: 12 Jul 2011, 9:33:12 UTC

CPU however, is set to 'use at most 25% of multiprocessors'

Which, on a Core 2 Duo means that BOINC should use no CPU at all. The value here is integer, meaning that if you set it to 99% on a 2 core CPU, that only 1 core will be used (or any value from 50 - 99% for that matter).
0 - 49% will tell BOINC to use zero cores, since INT anything less than half of a 2 core CPU is how much? Yep, nothing. ;-)

and 'use at most 30% of CPU time'

This will run science applications for 3 seconds out of every 10 seconds. The app will run at full speed (100% CPU), before being suspended. No, BOINC does not use any sustained values, where you will see in Windows Task Manager that it uses only 30% of a core, as showing that value needs to call a specific Windows API.

Since BOINC is a one source -> multi-platform (Ubuntu Linux, Windows and OS X) program, it cannot fall back for specific things on specific platform calls, as then that call will not work under the other platforms, without needing to release different source codes... So the developers opted for running and suspending of the science application, so that this method of 'throttling' will work under any platform.

And if you then want to use something more OS specific, use a third party application, such as there are for Windows Threadmaster and Fred's TTrottle.

I don't know enough of OS X to say if there are similar programs for your Mac.
ID: 38886 · Report as offensive
Robster

Send message
Joined: 4 Jul 11
Posts: 21
United Kingdom
Message 38896 - Posted: 12 Jul 2011, 6:28:00 UTC - in response to Message 38886.  

Hmm

Interesting.

It is a Core 2 Duo, the setting I gave are correct, so you are saying nothing should be happening, yet, BOINC is clearly working and taking a chunk of both cores!

Now I understand how BOINC throttles it makes sense why I see peeks and troughs on CPU usage.

EXCEPT under weatherprediction that was constantly taking 80% CPU!

Robin



ID: 38896 · Report as offensive
Profile Gundolf Jahn

Send message
Joined: 20 Dec 07
Posts: 1069
Germany
Message 38898 - Posted: 12 Jul 2011, 9:24:46 UTC - in response to Message 38897.  
Last modified: 12 Jul 2011, 9:27:01 UTC

That's a mystery to me.

I presume it's a local versus online preferences issue. ;-)

Gruß,
Gundolf
ID: 38898 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 38899 - Posted: 12 Jul 2011, 12:31:44 UTC

Look, previously, on BOINC 5 versions, this preference said "On multiprocessors, use at most X processors", where X would in your case be 1 or 2.

This preference changed to an integer percentage usage for BOINC 6, where you fill in how many CPUs or CPU cores you want to use. 25% of 2 cores is INT(half a core) == zero cores. I can't help it, it's what the software says must be happening.

So then we touch on a good thing, that Gundolf said: Where did you set the preferences? On a project's website->Computing preferences? Or in 6.10 BOINC Manager->Advanced->Preferences, or 6.12 BOINC Manager->Tools->Computing preferences?

If on the web site, then do check that you're not inadvertently using the local preferences instead. These override all the same web-preferences. You can then change preferences on the web site until your walls fall down from neglected maintenance, but they won't be used.

How to check that you're using the local preferences?
Two ways for that. The simplest really is to exit BOINC fully and restart it, then check in the Messages tab (6.10) or the Event Log (6.12) for a line that says "Date Time | | Reading preferences override file". If that line is there, you're using local preferences.

If you do not want to restart BOINC for some reason, the second way to check for this file is to navigate to your BOINC Data directory (by default that's in /Library/Applications Support/BOINC/ on OS X) and check for a file called "global_prefs_override.xml". If that file is in your Data directory, then you're using the local preferences.

And if that file is there, please post its contents here. You can open it with any simple text editor, there's no need to get an XML reader.

If that file is there, your next decision should be what you want to use. Local preferences or web-preferences? If web-preferences, then open the local preferences and click Clear. If you prefer the local preferences, then you should change your settings here.
ID: 38899 · Report as offensive
Robster

Send message
Joined: 4 Jul 11
Posts: 21
United Kingdom
Message 38912 - Posted: 12 Jul 2011, 20:51:18 UTC - in response to Message 38899.  

Hi

I have only ever set preferences locally.

I am certainly crunching data as I can see the project progress bars!

Here is the contents of the file:-

<global_preferences>
<run_on_batteries>0</run_on_batteries>
<run_if_user_active>1</run_if_user_active>
<run_gpu_if_user_active>0</run_gpu_if_user_active>
<idle_time_to_run>1.000000</idle_time_to_run>
<suspend_cpu_usage>45.000000</suspend_cpu_usage>
<start_hour>0.000000</start_hour>
<end_hour>0.000000</end_hour>
<net_start_hour>0.000000</net_start_hour>
<net_end_hour>0.000000</net_end_hour>
<leave_apps_in_memory>0</leave_apps_in_memory>
<confirm_before_connecting>0</confirm_before_connecting>
<hangup_if_dialed>1</hangup_if_dialed>
<dont_verify_images>0</dont_verify_images>
<work_buf_min_days>0.100000</work_buf_min_days>
<work_buf_additional_days>0.250000</work_buf_additional_days>
<max_ncpus_pct>25.000000</max_ncpus_pct>
<cpu_scheduling_period_minutes>60.000000</
cpu_scheduling_period_minutes>
<disk_interval>60.000000</disk_interval>
<disk_max_used_gb>20.000000</disk_max_used_gb>
<disk_max_used_pct>10.000000</disk_max_used_pct>
<disk_min_free_gb>20.000000</disk_min_free_gb>
<vm_max_used_pct>10.000000</vm_max_used_pct>
<ram_max_used_busy_pct>10.000000</ram_max_used_busy_pct>
<ram_max_used_idle_pct>90.000000</ram_max_used_idle_pct>
<max_bytes_sec_up>0.000000</max_bytes_sec_up>
<max_bytes_sec_down>0.000000</max_bytes_sec_down>
<cpu_usage_limit>35.000000</cpu_usage_limit>
<daily_xfer_limit_mb>0.000000</daily_xfer_limit_mb>
<daily_xfer_period_days>0</daily_xfer_period_days>
</global_preferences>

Robin
ID: 38912 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 38914 - Posted: 12 Jul 2011, 21:58:05 UTC - in response to Message 38913.  
Last modified: 12 Jul 2011, 22:02:39 UTC

<edit>

Well it looks like we've uncovered a bug here. I just set my 2 core system to use 25% of processors. In the messages it says:

Number of usable CPUs has changed from 2 to 1.
max CPUs used: 1


And sure enough, 1 project is still crunching away. That's on Linux, BOINC 6.12.33.

</edit>

Or it's a feature. Wouldn't surprise me if they changed it, as it would make more sense that you always use at least one CPU/core, instead of none. I'll go ask.

[edit] he he, when I change it to 0% it runs two cores on my 4 core. 1% = 1 core. ;-)
[/edit]
ID: 38914 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 38919 - Posted: 13 Jul 2011, 2:21:28 UTC - in response to Message 38916.  

It's a funny one... has nothing so much to do with the OP's problem, but it's something that best be written out somewhere, as you'll see someone in the near future needs it.

The old setting of "On multiprocessors, use at most X processors" is apparently still used in BOINC 6, as a background check for things like this. Mine was set to use 2 CPUs, which is why, when I told BOINC to use zero percent of the multiprocessors, it went and used 100% of the amount of CPUs I had told it to use in the old preference.

So Dag, in your case it probably says 8 there. Which is why, when you set it to 0%, it defaulted back to those 8 CPUs/cores.

There was a code change for 6.10 already, which said:
    if (global_prefs.max_ncpus_pct) {
        ncpus = (int)((ncpus * global_prefs.max_ncpus_pct)/100);
        if (ncpus == 0) ncpus = 1;
    } else if (global_prefs.max_ncpus && global_prefs.max_ncpus < ncpus) {
        ncpus = global_prefs.max_ncpus;
    }

The line "if (ncpus == 0) ncpus = 1;" was added, that made all the difference. Oh well, even I am still learning, every day. :)
ID: 38919 · Report as offensive
Robster

Send message
Joined: 4 Jul 11
Posts: 21
United Kingdom
Message 38946 - Posted: 14 Jul 2011, 12:46:35 UTC

Hmmmm

Very interesting to have helped identify a bug HOWEVER, this has not got to the bottom of the problem I had with climatepredition where it was taking a CONSTANT 80% CPU utilisation, not intermittent, not up and down but constant.

Interestingly and confusingly since the problem arose BOINC has not downloaded a single climatepredition task!

The only tasks I now have running are Seti and rosetta although climatepredition is still in the list of projects!

Robin
ID: 38946 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 38947 - Posted: 14 Jul 2011, 15:13:24 UTC - in response to Message 38946.  

That's because CPDN has been down for the past couple of days. They have severe server problems they want to fix first, and therefore they'd taken down the whole climateapps2.oucs.ox.ac.uk domain.

I see it's up now, but that they still have problems, so I don't expect them to give out work at this time. As for your 80% CPU use on a CPDN model, when things hang that can happen. While exiting & restarting BOINC will normally fix this, we do need to ask you to report this problem on their forums, since it was their application and probably their model that hung.
ID: 38947 · Report as offensive

Message boards : Questions and problems : HELP PLEASE Project NOT obeying preferences!

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.