log in
1) Message boards : Number crunching : Native R 2.15.1 Application (Cross Platform) v1.10 (sse2) task with Zero run time, Zero CPU time but valid? (Message 3322)
Posted 12 Nov 2014 by Profile Tom
The workunit was deleted form our database so I can't look at it directly. After 1-2 hours we checkpoint and return whatever we can back (hence the valid result). With zero runtime, CPU, etc. I suspect the process ran out of memory (or another task hogged all the memory from it).
Additionally, the current batch of R tasks are expected to run for at least an hour for a single simulation, so the progress bar will jump from ~1-2% to 100% when the simulation ends.

In any case, I looked at your recent work and didn't see this happen again, so we should be ok to write this off as an anomaly.

2) Message boards : Number crunching : Every wu is failing with Computation Error in Linux Mint (Message 3316)
Posted 6 Nov 2014 by Profile Tom
See https://mindmodeling.org/forum_thread.php?id=979&postid=3315.
3) Message boards : Number crunching : invalid signature (Message 3315)
Posted 6 Nov 2014 by Profile Tom
No, we updated some packages on our server, including Apache, and that caused some of our Apache configurations file to be overwritten. The BOINC daemons were temporarily disabled as we worked through these issues, but they should be online and working now. Please let me know if the issue persists.

Cheers,
Tom
4) Message boards : Number crunching : Every wu is failing with Computation Error in Linux Mint (Message 3311)
Posted 6 Nov 2014 by Profile Tom
Excellent! We don't currently have any Java tasks in the system, but the permissions issue may have affected the other apps as well. Hopefully, you're solution fixed the problem. Please keep me posted :)
5) Message boards : Number crunching : Every wu is failing with Computation Error in Linux Mint (Message 3309)
Posted 4 Nov 2014 by Profile Tom
Hi exalpha,

The Java application is getting a permission error when trying to run:

Could not start /home/greg/BOINC/projects/mindmodeling.org/java-1.7_1.0_env/java-1.7/bin/java execv() failed: : Permission denied


Could you check the permissions on this path:

home/greg/BOINC/projects/mindmodeling.org/java-1.7_1.0_env/java-1.7/bin/

6) Message boards : Number crunching : Native Java 1.7 Application requires 1.97 KB/sec download bandwidth... (Message 3308)
Posted 4 Nov 2014 by Profile Tom
Hi all,

Our workunits are typically scheduled to run an hour. When the system detects that results are not returning, it will reduce the size of any newly created workunits to a more manageable amount. Unfortunately, the work was returning from the clients fine, but the validator was broken. The work generator, not detecting this, continued to reduce the size of the workunits until they reached a threshold of only a few seconds. This lead to all the follow-on issues described above, which can be attributed to our Validator not running. It has since been fixed and workunits are processing again (albeit slowly).

I hope that helps explain the situation. You may need to reset (or detach/reattach) your project to get it back into a normal state again.
7) Message boards : Number crunching : Native Pypy version 1.9 workunits ending in computational error (Message 3303)
Posted 3 Nov 2014 by Profile Tom
Ok, the latest version ran without error on Windows hosts. Please let me know if you are still experiencing the problem-- after you reattach ;)

Cheers,
Tom
8) Message boards : Number crunching : Every wu is failing with Computation Error in Linux Mint (Message 3302)
Posted 3 Nov 2014 by Profile Tom
Hi Yacob,

The error your getting is different than the one outlined above. I'm testing a fix on the Windows platform as we speak.
9) Message boards : Number crunching : Native Pypy version 1.9 workunits ending in computational error (Message 3301)
Posted 3 Nov 2014 by Profile Tom
Hi Yacob,

I apologize for getting back to you so late. The problem is that the the Pypy version on Windows behaves a little differently than other operating system. I'm testing a fix right now, and will get to you shortly.

Cheers,
10) Message boards : Number crunching : Native Pypy version 1.9 workunits ending in computational error (Message 3291)
Posted 9 Oct 2014 by Profile Tom
Thanks everyone, I was able to track down the error this time. The Pypy application had some deprecated code that was loading in a module not available in the Windows libraries. This code has been removed and should now work on Windows hosts.
11) Message boards : Number crunching : Every wu is failing with Computation Error in Linux Mint (Message 3288)
Posted 8 Oct 2014 by Profile Tom
The Pypy task should be cross-platform. Our database purge operation deleted the failed workunits before I could inspect the error, so I can't inspect the problem anymore. I ran a Linux-only python model on 10/2/2014 with success, so the problem maybe be Linux distro specific.

I'll keep a lookout for errors the next time we run a Pypy model.
12) Message boards : Number crunching : Website error messages (Message 3282)
Posted 2 Sep 2014 by Profile Tom
Some of the BOINC code isn't properly checking for null objects, but there shouldn't be any adverse affects. I'm redirecting the warnings to sysadmin logs for now, but please let me know if you experience any unexpected results in the future.
13) Message boards : Number crunching : Warning Message (Message 3281)
Posted 2 Sep 2014 by Profile Tom
Thank you for reporting this bug. Although some of the BOINC scripts aren't properly checking for null objects, there doesn't seem to be any adverse consequence of this (other than the warning messages). Either way, I don't like error messages being reported on public facing pages, so I'm redirecting them to system admin logs.

Thank you!
14) Message boards : Cafe : DC-Vault Ongoing Challenge for Multiple Projects (Message 3280)
Posted 2 Sep 2014 by Profile Tom
Thanks Coleslaw :)
15) Message boards : Number crunching : Who many WU´s are in each batch (Message 3279)
Posted 2 Sep 2014 by Profile Tom
Yes, except I'll need to report simulations instead of workunits since we batch workunits dynamically at runtime.
16) Message boards : Number crunching : Stats export not working? (Message 3278)
Posted 2 Sep 2014 by Profile Tom
Excellent! Please let me know if this changes at any point in the future. We usually see this when we push out new code and permissions accidentally flip.
17) Message boards : Number crunching : Request: Increase the Deadlines! (Message 3277)
Posted 2 Sep 2014 by Profile Tom
Resetting the project will reset the average download calculations. We've seen this on some of our local systems. It may be a consequence of highs and lows in our work queues.
18) Message boards : Number crunching : Work for Windows? (Message 3276)
Posted 2 Sep 2014 by Profile Tom
Thanks for responding Conan. You are 100% correct.

For interpreted languages like Python or Lisp, or a language like Java that uses a virtual machine, we are able to execute the model on all platforms. But for compiled models written in C/C++, we are limited to running it on the platforms provided by the researcher.

We recently added a new research contributor to the project whom uses the EPIC cognitive architecture to explore models. It is written in C/C++ and compiled on the Linux platform. These are the jobs in the work queue as of late.

More information here:
http://en.wikipedia.org/wiki/EPIC_(cognitive_architecture)
19) Message boards : Number crunching : task aborted with "computation error" right after download (Message 3247)
Posted 14 Jul 2014 by Profile Tom
That is correct.
20) Message boards : Number crunching : "Project web pages" links (in BOINC Manager) are not correct (Message 3241)
Posted 9 Jul 2014 by Profile Tom
Fixed. I updated the Master URL in our project config file. If you're still experiencing issues, try resetting the project or detaching->attaching to download the updated config.


Next 20

Main page · Your account · Message boards


Copyright © 2020 MindModeling.org