| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Boinc on FreeBSD: exceeded disk limit |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 6
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
My FreeBSD 6.1 Release machine has been running fine with Boinc v5.4.9_1. Work units from all the projects have been producing valid results.
Since I updated Boinc to v5.8.11 only the Genome work units have been producing valid results. The HDC and FAAH work units have been failing to complete due to "exceeded disk limit" errors at approximately one hour of processing the work unit. In checking the results web page, it shows the other machines working on the same work units are producing valid results. The only change to the system was updating Boinc from v5.4.9_1 to v5.8.11 (and the dependancies). The amount of disk space available to Boinc has not changed. The amount of the disk limit varies from 476MB to as little as 74MB. This appears to be a limit imposed by the work unit. Though I have increased the amount of free space on this drive to no affect. The only thing showing up in stderrdae.txt, and not every time is: SIGPIPE: write on a pipe with no reader The working directory clears just as fast as the error occurs so the only items I have found in stderr.txt are: About to call graphics init dlopen() failed: libGL.so.1: cannot open shared object file: No such file or directory No graphics. [DIAG] Crop rect (T,L - B,R): 220, 212 - 1199, 1193 The above repeats a number of times before the work unit fails. And yes, libGL.so.1 is in the path. I suspect the "shared object file" is something the work unit should be producing. The most detailed information about an error I've seen in stdoutdae.txt is: 2007-02-21 22:14:02 [World Community Grid] Aborting task B24071_0047_YTMA89-32-2-9-c1_0: exceeded disk limit: 481.20MB > 476.84MB 2007-02-21 22:14:02 [World Community Grid] Deferring communication for 1 min 0 sec 2007-02-21 22:14:02 [World Community Grid] Reason: Unrecoverable error for result B24071_0047_YTMA89-32-2-9-c1_0 (Maximum disk usage exceeded) B24071_0047_YTMA89-32-2-9-c1.pgm wcg_tma_imagein.fvm 857234, 1440000 Bulid texton library... set 0 KNN: 75 clusters Other errors are as short as: 2007-02-21 12:03:30 [World Community Grid] Aborting task faah1358_d097n645_x2BPW_01_0: exceeded disk limit: 75.62MB > 71.53MB 2007-02-21 12:03:30 [World Community Grid] Deferring communication for 1 min 0 sec 2007-02-21 12:03:30 [World Community Grid] Reason: Unrecoverable error for result faah1358_d097n645_x2BPW_01_0 (Maximum disk usage exceeded) 2007-02-21 12:03:33 [World Community Grid] Computation for task faah1358_d097n645_x2BPW_01_0 finished Based on another thread on this error I have tried increasing the CPU usage limit, and setting "Leave applications in memory while preempted?" to "Yes" to no affect. Unless someone has a suggestion, I'm going to reinstall Boinc v5.4.9_1 until another update comes out. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Seen several commentaries about 'pipe' for a week or so on BOINC dev and some correction(s) having been checked in. Go to this search link http://boinc.berkeley.edu/dev/forum_search.php and type the SIGPIPE in the top box to get some discussion in the "BOINC 5.8 released to public" thread.
----------------------------------------The versions have been increasing rapidly to 5.8.15, so u might give that a try. Get it from the official site: http://boinc.berkeley.edu/download_all.php edit: storing searches does not work on BOINC dev, so changed instruction!
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Feb 22, 2007 9:56:18 AM] |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
Are you using the Throttle?
There appears to be a bug in the interaction between the throttle on the 5.8 client and checkpointing on the various unix like platforms (Linux, FreeBSD, Mac OS X). We are investigating the bug now. In the meantime, try setting the throttle back up to 100% and see if the errors go away. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Are you using the Throttle? Throttle? Do you mean the "Use no more than: % of processor time " setting in the device profiles? I did increase this from the default of 60% to 75%. Guess I could try 100% and see how it works. The pipe issue reported on the Boinc Dev forum appears to have no relation to the problem I'm seeing. I would like the newer versions to stablize before trying one. Right now I'm seeing a number of warnings of problems with the newer versions. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Versions of BOINC prior to 5.8 didn't use the throttle setting - they always used 100% of processor time.
You can downgrade BOINC to 5.4 or try setting the throttle to 100% to see what happens. The throttle has caused quite a few difficulties in BOINC 5.8. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
After changing the "Use no more than: % of processor time " setting in the device profiles to 100.0%, I now have one valid result with another pending.
 Thanks for everyone's help! Glad to have a workaround that keeps me from having to install another version. |
||
|
|
|