| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Everything since Feb either "Error" or too late |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 4
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Just noticed my average points had dropped, and now I see that as far back as I can see, all of my projects are either listed as "error" "too late" or "In progress" (may have a "pending" as well). I can only see back to about the 25th of Feb on this system, and the "Error" filter in the history returns no results. I only see "error" when I turn all filters off in the "results status" page.
I have not had any issues that I know of prior to this. Of course, it appears the system is not telling me that it has failed unless I log into the website and check results status. I did upgrade Boinc to the 7.2.39 (latest) version within the last few weeks. System is I7 core running Windows 7, X64, lots of RAM, not pushing any limits that I can see. No other projects are showing any issues (GPUGRID, SETI and Milkyway) Note that I am configured for all projects, though it appears that the only two that are not "In progress" and showing errors are FAHV, FAAH, MCM1 projects. Hate wasting processor cycles if they are just getting dumped. Have set the system to only process WCG, but not get new tasks until I clear the queue and get this sorted out. Any suggestions on how to fix this? Thanks! -Mike |
||
|
|
jonnieb-uk
Ace Cruncher England Joined: Nov 30, 2011 Post Count: 6105 Status: Offline Project Badges: 
|
In Result Status locate a task that has Error under Status. Click on Error. Copy and paste the entire report which will look something like this:
----------------------------------------Result Log Result Name: MCM1_ 0002743_ 6059_ 1-- <core_client_version>7.2.33</core_client_version> <![CDATA[ <stderr_txt> Commandline = projects/www.worldcommunitygrid.org/wcgrid_mcm1_7.28_windows_x86_64 -SettingsFile MCM1_0002743_6059.txt -DatabaseFile dataset-17_72_SDG_v1.txt Settings File DateOfDesign = 11/07/2013 Designer = PMCC_OCI WorkOrderID = 2743_6059 DatasetID = 17_72_SDG_v1 NumberOfGenesInStartingSignature = 13 NumberOfGenesInSignatureMin = 10 That's the first step in getting assistance,. If both MCM & FAAH are erroring out post example for each. ---------------------------------------- [Edit 1 times, last edit by jonnieb-uk at Mar 9, 2014 12:21:44 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks! I had spot checked the errors, but most are along the lines of:
----------------- Result Log Result Name: MCM1_ 0002587_ 3514_ 0-- <core_client_version>7.2.39</core_client_version ----end of error log---- Another: Result Log Result Name: FAHV_ x3NF9_ A_ IN_ Y3a_ rig_ 0199245_ 0851_ 0-- <core_client_version>7.2.39</core_client_version> ------end of log----- One I just found: Result Log Result Name: MCM1_ 0002611_ 8280_ 1-- <core_client_version>6.10.58</core_client_version> <![CDATA[ <message> too many exit(0)s </message> ]]> -----------------end log-------------- I see that some are now validating again. Thanks! -Mike |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Well, since I like to see the resolution when I am searching...
I believe I tracked this to my TThrottle utility settings. I had been working with them at some point, and had apparently set the max CPU temp down quite a bit lower than I should. Since this primarily impacted Boinc projects, it dropped the CPU activity for each project down to about 1/3 or less of normal. Add to this that WCG tends to have some short "complete by" dates compared to SETI and some of the others I support, and WCG tasks don't seem to upgrade to priority/urgent status even when they are about to expire. Even though the Resource Share was set to 400 (more than all other projects combined) I guess the algorithm was not detecting the reduced CPU cycles it was taking, so it (or perhaps the other projects) were sending a LOT more work than could be done by deadline. I think the WCG stuff was basically timing out. With TThrottle set properly, I am now getting them out on time (roughly per the estimated time to complete) and the results are validating. There is a bit of a race going on to get caught up, since I only finally figured this out a few days ago and a lot of work was queued up (stopped updates until I am caught up). So, basically, it was the much lower CPU availability/cycles caused by TThrottle being set too "cool", combined with quirks of the work estimating algorithms of WCG/Boinc and the other projects. I figured it out mostly by looking at the elapsed versus remaining times in the "TASK" tab , and taskmanager CPU load by task. I also wondered why even when I set Boinc to run 100% of my CPU, there was no slowdown in other programs (TThrottle only impacts Boinc tasks, not the others). Hope this helps someone else... (or me, when I forget 6 months from now and do it again). -Mike |
||
|
|
|