Message boards : Number crunching : Can't get tasks or reset project
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 37971 - Posted: 11 Feb 2019, 2:21:45 UTC
Last modified: 11 Feb 2019, 2:31:38 UTC

My last theory task resulted in computation error because of a lack of memory (I've since ordered more DRAM to double my 6GB). I've been unable to download a new task. I reset the project, since I had nothing to lose. Last time I did that it resulted in a long download. This time no download, and the Program Data\BOINC\Projects\lhcathome folder is empty. I even removed the project and rejoined. I tried downloading every application, but no luck. I don't know what to make of it, or what to do.

Edit: I forgot to mention that I'm also running two Einstein@Home GPU tasks. Each uses 1 CPU of my quadcore, and has a working set of ~168 MB in system memory. They each consume 25% of total processor capability.
ID: 37971 · Report as offensive     Reply Quote
Jonathan

Send message
Joined: 25 Sep 17
Posts: 93
Credit: 3,078,808
RAC: 2,719
Message 37972 - Posted: 11 Feb 2019, 2:54:55 UTC - in response to Message 37971.  

Can you look in the Boinc Manager Event log and see what is reported?
ID: 37972 · Report as offensive     Reply Quote
Profile Ray Murray
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 281
Credit: 11,859,285
RAC: 1
Message 37977 - Posted: 11 Feb 2019, 19:18:23 UTC - in response to Message 37971.  

Hi, Steve,
The Exceeded Memory Limit error isn't your, or your host's, fault. Some Sherpa jobs have a fault that makes them write a log that is FAR too big, in the GB range rather than MB or even kB and gets larger than the limit set at Project level. I see you've been unlucky in getting 5 of them in the last week 8¬(
These errors and the Aborts count against you in Task allocation with each failure reducing the number of tasks you are permitted (it's a safeguard so that rogue hosts don't chew up all the work) so it might be that you have to wait until the allocation resets (23:00 UTC, I think?) There will be a message in Event Log if you are currently not being sent work for that reason.

I can't remember how long Einsteins run so it could just be that your job cache is full? Maybe try suspending any waiting Einsteins and see if that will bring in a Theory.

Hope that helps. If not, come back and let us know.
ID: 37977 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 15 Jun 08
Posts: 2386
Credit: 222,990,641
RAC: 136,455
Message 37978 - Posted: 11 Feb 2019, 19:56:26 UTC - in response to Message 37977.  

... These errors and the Aborts count against you in Task allocation with each failure reducing the number of tasks you are permitted (it's a safeguard so that rogue hosts don't chew up all the work) ...

Very unlikely that this is the reason for the problem as the number of tasks is not high enough.
See:
https://lhcathome.cern.ch/lhcathome/host_app_versions.php?hostid=10582787


... it could just be that your job cache is full? Maybe try suspending any waiting Einsteins and see if that will bring in a Theory ...

More likely than the above.
I would try the suggested method.
ID: 37978 · Report as offensive     Reply Quote
Profile Ray Murray
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 281
Credit: 11,859,285
RAC: 1
Message 37979 - Posted: 11 Feb 2019, 20:58:20 UTC - in response to Message 37978.  
Last modified: 11 Feb 2019, 20:59:01 UTC

Thanks Computezrmle.
That's the link I was looking for before I made the earlier post but couldn't remember where to find it so just put it out as a possible reason, which now having that info, clearly this isn't the cause.
ID: 37979 · Report as offensive     Reply Quote
Profile Magic Quantum Mechanic
Avatar

Send message
Joined: 24 Oct 04
Posts: 1114
Credit: 49,503,137
RAC: 3,956
Message 37981 - Posted: 11 Feb 2019, 21:42:55 UTC

https://lhcathome.cern.ch/lhcathome/result.php?resultid=215698053

As I mentioned before you have to set your Boinc Manager Options- computer preferences-disk and memory settings so that you allow it to use your memory..........I have all 10 of my computers set at 100%

This means it is available NOT that it just uses all of that memory all the time.

And you are set to run all 4 CPU's in your LHC preferences so you were running a 4 CPU task with not enough Memory.

Einstein GPU's are not the problem........it uses memory from your GeForce 750 card.

I have been running both for the last 8 years and a pc similar to yours with only 12GB ram running 2 GPU tasks and 2 of the 2-core Theory tasks 24/7 and they never fail set with 100% memory available.

I have the CPU's set at 100% also but as I said that does not mean these Theory tasks will use 100% but it sure does make these tasks run without Invalid tasks.........I sort of prefer Valid ones

As far as reinstalling Boinc and VB I NEVER do both from here.......always do the VB separate from the d/l site and then reboot......never fails.
https://www.virtualbox.org/wiki/Download_Old_Builds_5_2
ID: 37981 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 37983 - Posted: 12 Feb 2019, 0:41:42 UTC - in response to Message 37981.  

https://lhcathome.cern.ch/lhcathome/result.php?resultid=215698053

As you can see from the ratio of CPU time to Run time of ~2 I was only using 2 CPUs when this failure occured. When I was using 4 CPUs I wasn't running Einstein, just Theory.

you have to set your Boinc Manager Options- computer preferences-disk and memory settings so that you allow it to use your memory

Per your advice I have changed memory usage to 100%, but I did it on the website, not in BOINC.
And you are set to run all 4 CPU's in your LHC preferences so you were running a 4 CPU task with not enough Memory.

Are you saying that running a 4 CPU task uses more memory than a 2 CPU task? I thought it just meant that the task of a given size was divided up into however many CPUs are available?
Einstein GPU's are not the problem........it uses memory from your GeForce 750 card.

Einstein tasks do use a small amount of system memory...~158 MB if memory serves, as well as virtual memory.
I have been running both for the last 8 years and a pc similar to yours with only 12GB ram running 2 GPU tasks and 2 of the 2-core Theory tasks 24/7 and they never fail set with 100% memory available.

I appreciate all the help you gave me in PMs.
ID: 37983 · Report as offensive     Reply Quote
bronco

Send message
Joined: 13 Apr 18
Posts: 443
Credit: 8,438,885
RAC: 0
Message 37984 - Posted: 12 Feb 2019, 0:52:19 UTC - in response to Message 37983.  

Are you saying that running a 4 CPU task uses more memory than a 2 CPU task?

Yes. But not twice as much, less than twice as much. There was a thread around here somewhere about a month ago that discusses the formulae for calculating RAM requirement for all the various apps.
ID: 37984 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 37985 - Posted: 12 Feb 2019, 0:53:53 UTC - in response to Message 37972.  
Last modified: 12 Feb 2019, 1:00:04 UTC

Sorry, Jonathan,. When I first accessed BOINC Event Log it had data from 2/10/19 US Pacific time. But I've since rebooted the system several times, and now it only covers today starting at 3:56 PM. Does the log reset whenever BOINC is shut down?

Edit: The first Event Log did say that there were Theory tasks for ATI GPUs, but my preferences did not indicate I wanted that. I've since corrected that. Later entries all said that there are no Theory WUs.
ID: 37985 · Report as offensive     Reply Quote
bronco

Send message
Joined: 13 Apr 18
Posts: 443
Credit: 8,438,885
RAC: 0
Message 37986 - Posted: 12 Feb 2019, 1:11:26 UTC - in response to Message 37985.  

What you see in the Event Log in the manager resets every time you start BOINC client. Log entries from the previous session might appear in the log saved on disk. That log might go back several days. You can view that log by opening it in a text editor, Windows Notepad for example. Do NOT open it in a word processor such as MS Word.

There is a setting that allows you to control how many entries are retained in the log. Think it's in your personal prefs?
ID: 37986 · Report as offensive     Reply Quote
bronco

Send message
Joined: 13 Apr 18
Posts: 443
Credit: 8,438,885
RAC: 0
Message 37987 - Posted: 12 Feb 2019, 1:17:45 UTC

Link to that thread with the formulae for RAM requirements for LHC tasks... https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4875
ID: 37987 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 37988 - Posted: 12 Feb 2019, 2:07:42 UTC - in response to Message 37987.  

Link to that thread with the formulae for RAM requirements for LHC tasks... https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4875

Thanks bronco. I was also interested in Sixtrack, but that's not included in the link. Have any idea about that?
Also, when I access the project preferences ALICE and LHCb have disappeared from the list. At first I thought it might be because in my quandary I tried to download everything, but they had disappeared because I don't have enough memory yet. But then I realized that since ATLAS is well known for needing a lot of memory, and it's still on the list, that couldn't be the reason. Maybe there just aren't any tasks now or expected in the near future for those applications.
ID: 37988 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 37989 - Posted: 12 Feb 2019, 2:14:20 UTC - in response to Message 37986.  

Log entries from the previous session might appear in the log saved on disk. That log might go back several days.

I can't find it in Program Files\BOINC. Can you tell me where it is?
There is a setting that allows you to control how many entries are retained in the log. Think it's in your personal prefs?

I can't find that either. I started out with SETI years ago. One would think I might know more about it.
ID: 37989 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 37990 - Posted: 12 Feb 2019, 4:01:51 UTC - in response to Message 37971.  
Last modified: 12 Feb 2019, 4:15:38 UTC

I wanted to thank everybody for all the help. At ~00:15 UTC I tried to download a 2 CPU Theory task and it worked, using 50%+ of total processor capability (I have a core-i5, meaning 4 cores).

Then I downloaded two Einstein tasks. In Einstein preferences I specified use at most 25% of processors, which feeds the GPU. So I expected to see 75% of total processor usage. But free memory at that time was just a few MB, although available memory was ~2900 MB. One of the Einstein tasks started, but the Theory task dropped, first to 25% then 0%, with Waiting to run. I scrambled to abort both Einstein tasks (I don't know why I didn't go for pause), and the Theory task started up again at 50%. I know LHC tasks don't like to be interrupted, so I guess until my additional 8 GB of memory arrives I'll forget about Einstein and use 3 CPUs for LHC. I used 4 cores for a couple of days, but this is the only computer I have, and I have to use it for other purposes.

"Remember, nothing that's good works by itself, just to please you. You have to make the damn thing work."
Thomas A. Edison
ID: 37990 · Report as offensive     Reply Quote
bronco

Send message
Joined: 13 Apr 18
Posts: 443
Credit: 8,438,885
RAC: 0
Message 37991 - Posted: 12 Feb 2019, 8:47:33 UTC - in response to Message 37988.  
Last modified: 12 Feb 2019, 8:49:22 UTC

Link to that thread with the formulae for RAM requirements for LHC tasks... https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4875

Thanks bronco. I was also interested in Sixtrack, but that's not included in the link. Have any idea about that?

It's there. Use your browser's search function(CTRL-F) to search for sixtrack.
Also, when I access the project preferences ALICE and LHCb have disappeared from the list.

Those two were taken out of service and sent off for repair because they were not working properly. There was mention of fixing them and returning them to service. Now that they're off the list it looks like they might not return any time soon.

The other questions you asked are best answered by a Windows user. I'm just Linux.
ID: 37991 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 37995 - Posted: 12 Feb 2019, 16:24:49 UTC - in response to Message 37991.  

It's there. Use your browser's search function(CTRL-F) to search for sixtrack.

Thanks again. I guess I glossed over those mentions of Sixtrack.
ID: 37995 · Report as offensive     Reply Quote
bronco

Send message
Joined: 13 Apr 18
Posts: 443
Credit: 8,438,885
RAC: 0
Message 38000 - Posted: 13 Feb 2019, 0:27:54 UTC - in response to Message 37995.  

It's easy to miss in such a long thread.
Regarding how many entrys are retained in Event Log, I was wrong, there is no clickable setting, it's configured in cc_config.xml. Details are in the official docs at https://boinc.berkeley.edu/wiki/Client_configuration, see

<max_event_log_lines>N</max_event_log_lines>
Maximum number of lines to display in BOINC Manager's Event Log window (default 2000, 0 means no limit). List-add.pngNew in 7.1.2
<max_file_xfers>N</max_file_xfers>
Maximum number of simultaneous file transfers (default 8).
<max_file_xfers_per_project>N</max_file_xfers_per_project>
Maximum number of simultaneous file transfers per project (default 2).
<max_stderr_file_size>N</max_stderr_file_size>
Specify the maximum size of the standard error log file (stderrdae.txt); default is 2 MB.
<max_stdout_file_size>N</max_stdout_file_size>
Specify the maximum size of the standard out log file (stdoutdae.txt); default is 2 MB.
ID: 38000 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 38012 - Posted: 14 Feb 2019, 16:07:24 UTC - in response to Message 38000.  

Thanks for the link and the message bronco. I'll have to ponder that some more. :-)
ID: 38012 · Report as offensive     Reply Quote
Steveplanetary

Send message
Joined: 14 Dec 11
Posts: 23
Credit: 101,345
RAC: 0
Message 38031 - Posted: 19 Feb 2019, 2:19:00 UTC - in response to Message 37981.  

I have the CPU's set at 100% also but as I said that does not mean these Theory tasks will use 100% but it sure does make these tasks run without Invalid tasks

My problem now is not that I can't get tasks. Now Theory tasks always run at 25%, even when max # CPUs=3, and whenever I start an Einstein task Theory immediately goes into Waiting to run. I changed my memory settings to 100% earlier. And I already had LHC preferences set at use 100% of processors 100% of the time. But I just now changed Einstein preferences from use at most 25% of the processors 100% of the time to 100% of the processors 100% of the time. When I resumed E@H, no problem, they're both running!

I'll restore my LHC project preferences to use 3 CPUs, but I'm not expecting a new download until 2:30 AM, so I probably won't see what happens with Theory. I'll keep you posted.

Thanks!
ID: 38031 · Report as offensive     Reply Quote
mrchips

Send message
Joined: 16 May 14
Posts: 15
Credit: 7,343,729
RAC: 0
Message 38796 - Posted: 13 May 2019, 11:25:24 UTC

Can't report my finished tasks and request new WU
Getting this message SERVER ERROR: FEEDER NOT RUNNING
ID: 38796 · Report as offensive     Reply Quote
1 · 2 · Next

Message boards : Number crunching : Can't get tasks or reset project


©2024 CERN