Compute Error? What gives?
11 Nov 2008 20:23:34 UTC
Topic 194026
(moderation:
Hi.
I am participating since 4 days in this effort (Boinc, Einstein@home), because my box is doing nothing most of the time.
During these last four days the computer here was assigned 13 tasks. Of these 13 tasks one is pending, one is sucessfully completed, and 11 ended with a "compute error". Some of those terminated after a few minutes, some others after 12 hours or more.
11 out of 13 is about 84%. I have a hard time thinking of any other technical system with such a failure rate.
Is there anything I am doing wrong? Or is there something wrong with the software? Or the data fed to my box? I checked the forum here and the FAQ, to no avail. All WU were from S5R4 AFAIK, for example the currently pending task is h1_0670.75_S5R4__63_S5R4a_0
The computer is an AMD Athlon 2600, 2GB RAM, Gentoo Linux with kernel 2.26.24. Boinc Manager is 6.2.15. It runs fine otherwise without any errors for weeks and months.
Any idea what that is?
Regards,
Ekki/Germany
Compute Error? What gives?
)
Looking at your tasks, you get two different error messages. Both are in my FAQs. ;-)
Exit code 99 with "Input Domain Error" in the stderr output. Although it says in that FAQ that it was for the S5R2 run, it also applies to the present run.
Exit code 38, signal 8
It is very likely that this
)
It is very likely that this bug is caused by a certain bug in the Linux Kernel.
If you compiled your Kernel with CONFIG_PREEMPT=y , you will see this problem in older kernel versions.
There is no known workaround, the only fix is to use a kernel that is not compiled with this setting or already has the correction patch included.
CU
Bikeman
P.S.: The web-stats indicate you are using Kernel 2.6.24-gentoo-r8
RE: RE: Any idea what
)
Thanks for the reply to you and bikeman (sorry about the wrong kernel version).
Well, where do I find these error messages?
- In "BoincManager" I have a page "Messages", but there I dont find any Error numbers like 99 or 38. The only message looking like an error message is "Output file h1_0670.75_S5R4__61_S5R4a_0_0 for task h1_0670.75_S5R4__61_S5R4a_0 absent"
- The file stderrdae.txt has zero length...
So where do I see these error numbers?
Anyway, I will set CONFIG_PREEMPT to N and try again.
Rgds,
Ekki
RE: So where do I see these
)
You can view your tasks through your Tasks for computer page, which you can access through Your account page. From the "Task for computer" page you can access the info of each task through the link "Task ID click for details", such as this example of your latest work unit.
Ok, it seems to work now with
)
Ok, it seems to work now with a new kernel with CONFIG_PREEMPT unset. Two completed WUs in the last few hours. Thanks for the support.
Rgds,
Ekki