2 Client errors in a row
22 Aug 2006 15:42:21 UTC
Topic 191720
(moderation:
I've just had two client errors in a row on my iMac 333, OSX 10.3.9, running BOINC 5.4.9 and Einstein S5R1_4.12_powerpc-apple-darwin.
Here are the WUs:
12288058
12311959
These are the first two that my machine has done since Einstein changed the credit earned algorith. Did something else get changed to prevent me from being able to run Einstein?
C
[/url]
2 Client errors in a row
)
Both those exited with this error: Maximum CPU Time Exceeded
It's possible E@H set the maximum time too low on these results. Bruce, Bernd?
Talking about results returned erroneous, I see one of your other machines is only returning errors. Most, if not all (I didn't check them all), have exit error -108. That's a System FOpen error, usually caused by hiding the BOINC directory and all subdirectories. BOINC doesn't know where to write its state to, or read the state from. It's easily solved by making sure the BOINC directory and all subdirectories aren't hidden.
Thanks for the report. I
)
Thanks for the report. I think indeed that adjusting the credit in the current way may have caused this as an unwanted side effect on slow machines. We're looking into this, this is probably a problem of workunit generation.
BM
BM
Jorden: Thanks - I've been
)
Jorden:
Thanks - I've been wondering about that machine - it belongs to a friend in another state. I'll call him and have him delete the Boinc files. I think it must have something to do with his children having individual login accounts on the machine.
Bernd: I'll go ahead and suspend Einstein on that machine so I don't delay someone else getting credits.
C
Join Team MacNN
I'm having a similar problem:
)
I'm having a similar problem: I had three out of four WUs give client errors (Work Unit IDs 12675829, 12674986, and 12672861) and the fourth (Work Unit ID 12673550) took about 50% longer than normal (36,789 sec rather than 24,000 to 25,000 sec). I reset Einstein (after clearing out all tasks first) and started crunching again. Now, after 14 successful results, I just had a validate error (WU ID 12792049). Any ideas on what's going on?
Here's my system:
MacBook with 1.83 GHz CoreDuo processors
BOINC 5.4.9
Einstein S5R1 4.0.3
Thanks.
Jim Milks
The "validate error" result
)
The "validate error" result file looks pretty bad. It isn't correctly compressed, and of what can be uncompressed the first lines look ok, but the last lines look like complete trash of digits, dots, blanks and linebreaks. This might be the result of a bad compression.
There are many possible reasons for that. My first guess would be transmission problems, it may also be a problem with the harddisk, RAM or CPU, and may or may not be temperature dependent.
BM
BM
Well, I just reported my
)
Well, I just reported my first two tasks since the validate error file. It looks like both of them were transmitted correctly. So whatever went wrong with WU ID 12792049, it does not appear to be permanent. Strange--I've crunched ~56,000 credits worth of work units and just had my first four errors (three client errors, one validate error) within the last three days.
Jim Milks
Jim, do you know about the
)
Jim, do you know about the Mac Beta test applications that should help speed up your processing time? Bernd has made some great improvements over 4.03. Go to the Number Crunching Forum, and see the two Mac Beta Test sticky threads, and pursue the correct choice for you.
Thanks for the tip. I just
)
Thanks for the tip. I just downloaded the Intel Mac Beta S5R1 4.28. I'll have to see what happens with it. It seems that S5R1 4.03 may be getting flaky on my machine--I had another WU give a validation error last night (WU#12877134). That makes it two out of the six work units I reported yesterday. Strange thing is that I had two work units to report. One returned a validation error, the other (WU#12873155) was a success. Same thing happened with the previous time I got a validation error (I have a CoreDuo machine, so I always have two units to report at a time). And I just reported my first two WU's of the morning. Both were successfully transmitted. Weird.
Jim
RE: Bernd: I'll go ahead
)
Bruce fixed this today, the WUs that are now inserted in the DB shouldn't cause this problem anymore. However currently you will probably get a Task with old settings, so better wait a bit longer before resuming the project on this machine.
BM
BM
RE: RE: Bernd: I'll go
)
I also fixed this retroactively. Any WU that your machine downloads from the server should have a higher CPU time limit.
Bruce
Director, Einstein@Home