Computation Errors
log in

Advanced search

Message boards : Number crunching : Computation Errors

Author Message
crashtech
Send message
Joined: 10 Apr 19
Posts: 28
Credit: 466,952,134
RAC: 6,966,894
Message 5103 - Posted: 19 Apr 2019, 0:48:52 UTC

Hi,

I've been getting only computation errors on six machines in two locations. Clearly I am doing something wrong globally. Most of the errors look like this:

<core_client_version>7.12.1</core_client_version> <![CDATA[ <stderr_txt> 10:56:20 (229436): wrapper (7.5.26012): starting 10:56:20 (229436): wrapper: running llr.exe ( -d -oPgenInputFile=input.prp -oPgenOutputFile=primes.txt -oDiskWriteTime=10 -oOutputIterations=50000 -oResultsFileIterations=99999999 -t 4) 10:56:22 (229436): llr.exe exited; CPU time 0.000000 10:56:22 (229436): called boinc_finish(0) </stderr_txt> <message> upload failure: <file_xfer_error> <file_name>R879_350-400k_wu_6538_1_0</file_name> <error_code>-240 (stat() failed)</error_code> </file_xfer_error> </message> ]]>


Any help?

Profile PDW
Send message
Joined: 15 Oct 15
Posts: 41
Credit: 1,078,210,394
RAC: 43,439
Message 5104 - Posted: 19 Apr 2019, 6:53:04 UTC - in response to Message 5103.

Hi,

Have you tried using an app_config.xml file to limit the number of running tasks to 1 to start with ?

It is unlikely all your 100 processor machines can run up to 25 tasks at a time :)

Profile rebirther
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 2 Jan 13
Posts: 7227
Credit: 42,729,227
RAC: 34
Message 5105 - Posted: 19 Apr 2019, 7:07:03 UTC

your -t command is wrong, use -t4 not -t 4

crashtech
Send message
Joined: 10 Apr 19
Posts: 28
Credit: 466,952,134
RAC: 6,966,894
Message 5106 - Posted: 19 Apr 2019, 16:59:03 UTC - in response to Message 5105.

your -t command is wrong, use -t4 not -t 4


Thank you. That appears to have been the problem.

crashtech
Send message
Joined: 10 Apr 19
Posts: 28
Credit: 466,952,134
RAC: 6,966,894
Message 5107 - Posted: 19 Apr 2019, 22:04:34 UTC

Just a note, in my case, I ended up putting an "<avg_ncpus>x</avg_ncpus>" line (x being the same value as in the <cmdline> entry) to obtain the expected multithreaded behavior seen with other apps. Maybe this is wrong too, but it does seem to do what I want it to do now.

Profile PDW
Send message
Joined: 15 Oct 15
Posts: 41
Credit: 1,078,210,394
RAC: 43,439
Message 5108 - Posted: 20 Apr 2019, 5:43:16 UTC - in response to Message 5107.

That's what I had to do, Conan pointed it out here...

http://srbase.my-firewall.org/sr5/forum_thread.php?id=1072&postid=4773#4773

Profile marmot
Avatar
Send message
Joined: 17 Nov 16
Posts: 97
Credit: 126,410,450
RAC: 21,911
Message 5122 - Posted: 20 Apr 2019, 10:24:39 UTC - in response to Message 5104.
Last modified: 20 Apr 2019, 10:33:33 UTC

Hi,

Have you tried using an app_config.xml file to limit the number of running tasks to 1 to start with ?

It is unlikely all your 100 processor machines can run up to 25 tasks at a time :)


I know of one defunct project, that wasn't NCI, that would run 96 WU on a 16 thread computer, get credits and complete the WU's. Well, until they increased the RAM per WU and sent my machine into a swap crisis which required a hard reset.

EDIT: Oooo, crashtech has a E5-2689 0. Never seen one of those in BOINC computer databases, had to look it up. No AVX, benchmarks faster than the 2690 0 althhough it's base clock is 300mhz lower, and runs 20W cooler.

crashtech
Send message
Joined: 10 Apr 19
Posts: 28
Credit: 466,952,134
RAC: 6,966,894
Message 5124 - Posted: 20 Apr 2019, 14:04:43 UTC - in response to Message 5122.

Hi,

Have you tried using an app_config.xml file to limit the number of running tasks to 1 to start with ?

It is unlikely all your 100 processor machines can run up to 25 tasks at a time :)


I know of one defunct project, that wasn't NCI, that would run 96 WU on a 16 thread computer, get credits and complete the WU's. Well, until they increased the RAM per WU and sent my machine into a swap crisis which required a hard reset.

EDIT: Oooo, crashtech has a E5-2689 0. Never seen one of those in BOINC computer databases, had to look it up. No AVX, benchmarks faster than the 2690 0 althhough it's base clock is 300mhz lower, and runs 20W cooler.


I was not aware that the 2689 did not have AVX, would you mind linking to the source of the info? cpu-world says that it does:

http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon%20E5-2689.html

Also about the 100 CPU thing, I just do that for convenience, it makes it so the CPU Limits field in the BOINC app becomes simply ncpus instead of a percentage. It's also convenient for other things too on occasion.

Profile rebirther
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 2 Jan 13
Posts: 7227
Credit: 42,729,227
RAC: 34
Message 5134 - Posted: 21 Apr 2019, 10:31:42 UTC - in response to Message 5108.

That's what I had to do, Conan pointed it out here...

http://srbase.my-firewall.org/sr5/forum_thread.php?id=1072&postid=4773#4773


Added to the FAQ.

Profile marmot
Avatar
Send message
Joined: 17 Nov 16
Posts: 97
Credit: 126,410,450
RAC: 21,911
Message 5137 - Posted: 21 Apr 2019, 23:24:01 UTC - in response to Message 5124.
Last modified: 21 Apr 2019, 23:25:22 UTC



I was not aware that the 2689 did not have AVX, would you mind linking to the source of the info? cpu-world says that it does:

http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon%20E5-2689.html


From CPU Boss comparison to an E5-2690, it pops right out: http://cpuboss.com/cpus/Intel-Xeon-E5-2690-vs-Intel-Xeon-E5-2689

Although, instead of relying on spec sheets, that could be wrong, CPU-Z should report the abilities properly.
It's a potential upgrade to my current E5-2660's. Though, I'm more interested in performance/watt than maximum performance.

crashtech
Send message
Joined: 10 Apr 19
Posts: 28
Credit: 466,952,134
RAC: 6,966,894
Message 5144 - Posted: 24 Apr 2019, 4:01:05 UTC - in response to Message 5137.
Last modified: 24 Apr 2019, 4:02:53 UTC

From CPU Boss comparison to an E5-2690, it pops right out: http://cpuboss.com/cpus/Intel-Xeon-E5-2690-vs-Intel-Xeon-E5-2689

Although, instead of relying on spec sheets, that could be wrong, CPU-Z should report the abilities properly.
It's a potential upgrade to my current E5-2660's. Though, I'm more interested in performance/watt than maximum performance.

Good idea!


https://i.postimg.cc/9f0s3m8g/2689.jpg

I'm not sure if the img tag is working here?

My wish is to upgrade that platform this year, so perhaps I should keep you in mind when it's time to find another home for the CPUs.


Post to thread

Message boards : Number crunching : Computation Errors


Main page · Your account · Message boards


Copyright © 2014-2024 BOINC Confederation / rebirther