Task ends with computation error

The message is not about whether or not the PC owner is present or absent :-).

I've added some emphasis to the message to show up the real meaning. So you have to ask yourself, "What is the 'output file' and how could it be missing (ie absent)?"

Please realise that I'm just a volunteer like yourself so I certainly can't speak with the same assurance and level of accuracy as the person who actually wrote the code. The following comments may not be fully correct but do represent my understanding of the situation. Hopefully someone will correct me if my understanding is flawed.

When you request new GW tasks from the project, a lot of the time you will not need new data files since you will already have these. Each set of data files is potentially good for hundreds of tasks so a lot of the time, for a new task, all you will receive is a very small new 'result template' into which the results of the new calculation on the existing data can be written. I believe this result template is 'filled in' by the science app at the end of the run, using the output that accumulates during the full calculation. So it would be quite a problem if anything were to happen to the 'output file' at any stage during the run. If a hardware problem occurred that caused corruption of the output stream, the output file would be effectively 'lost'. So 'absent' may simply mean 'unable to be appended to or correctly interpreted due to some unknown hardware glitch'.

I have seen lots of examples of these over the years with my own hosts and I've always been able to find a likely hardware issue to explain them. Some of these include, excessive heat, excessive overclocking, faulty RAM, faulty PSU, faulty mobo capacitors, disk bad sectors, etc.

Is your machine a notebook? E@H tasks do put a heavy load on any computer and notebooks are particularly susceptible to thermal problems.

It is quite unlikely that the compute errors you are experiencing are in any way related to the version of BOINC you are using. It's much more likely to be a hardware issue with your machine. Try crunching with only one of the cores of your dual core. This would put a lot less heat stress on the machine. If the errors stop, you will know the problem is probably due to inefficient cooling.

Hi roghurt.

I have looked at your tasks and believe that I and a few others have had this same problem before.

If you look at the output of WUs you will see that the exit code is 38 and further down a message "process got signal 8" which is probably the real problem. Look at this thread and search for CONFIG_PREEMPT. Then check the preemption settings of your kernel by typing something like

cat /proc/config.gz | gzip -dc | grep PREEMPT

in a command shell. On my machine the output looks like this

# CONFIG_PREEMPT_RCU is not set
# CONFIG_PREEMPT_RCU_TRACE is not set
# CONFIG_PREEMPT_NONE is not set
CONFIG_PREEMPT_VOLUNTARY=y
# CONFIG_PREEMPT is not set

Notice the last line "# CONFIG_PREEMPT is not set". If your machine shows

CONFIG_PREEMPT=y

Then that is probably the problem. Re-compile the kernel with either

CONFIG_PREEMPT_VOLUNTARY=y

or

CONFIG_PREEMPT_NONE=y

That should fix it.

Good luck,
Michael

We can't view your results like that - only you can.

When you're in that list, click on the left-most link (under Task ID) of the work unit that got the error, then paste in that page's URL.

If you want it to be a clickable link, surround it with

[url] and [/url], like
http://full.URL.tld/etc

otherwise the forum software displays it as plain text and it must be copy/pasted to the browser's address bar (though there are add-ons that allow highlighting then right-clicking the highlighted text to open it 'hot' in another tab; not everyone has those installed).
e.g.
like this http://einsteinathome.org/task/200867643