All outcomes saying "client error"
1 Apr 2008 15:19:05 UTC
Topic 193603
(moderation:
Everything was working for a couple of days, with the outcomes all saying "success". Then, a couple of days ago, I started getting client errors on every single computation. Here is an example:
http://einsteinathome.org/task/94515676
They all have the same error:
*** glibc detected *** ../../projects/einstein.phys.uwm.edu/einstein_S5R3_4.38_i686-pc-linux-gnu_1: munmap_chunk(): invalid pointer: 0x08c11088 ***
Any ideas before I leave the project? Thanks.
All outcomes saying "client error"
)
Just for curiosity: What distro is this?
Please try to put a file named "CPU_TYPE_0" (without quotation marks) into your BOINC directory and see if the same thing happens with einstein_S5R3_4.38_i686-pc-linux-gnu_0.
BM
BM
RE: Just for curiosity:
)
The distro is Red Hat Advanced Server 5. I will try your suggestion when I get a chance.
RE: RE: Just for
)
No dice. Thanks for trying, though.
RE: No dice. Thanks for
)
Just out of curiosity, I had a look through your tasks list for that machine - actually the page showing the last successful results and the first of the errors. It was interesting to note that the first two tasks (dual core machine) showing errors also showed that crunching had been stopped approximately half way through in each case and then restarted after what appears to be quite a time lapse. For instance, the first error result was started at 04:43 UTC on March 28 and was restarted at 20:00 UTC at skypoint position 704/1202, ie. just over half way through. At that point about 4 hours of actual crunch time would have been accumulated so crunching had been stopped for about 11 hours or so. The other task seemingly shows a similar position.
I'm guessing that the machine may have been down for maintenance/upgrade of some sort and that perhaps new or updated software may have been installed? Is it possible that something happened at this time that is now having an unintended affect on BOINC? It doesn't seem to be anything to do with the science app as the crunching seems to be completing quite successfully in most cases.
Cheers,
Gary.
RE: RE: No dice. Thanks
)
Therefor i had made the decision to not change anything anymore on my cruching pc. I had make experiences with team-networking for load-balancing and fault-redundance or had make the OS-updates automatic in the background while boinc was also running and ended up with a screwed host and unexpected error 0x000005 on read request...
Murphy's Law is true and the often used words: "Never change a running system!" are gone into my blood.