| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Beta Testing
Forum: Beta Test Support Forum
Thread: New Beta Test for PC - September 18, 2015 [ Issues Thread ] |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 172
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
As you can see 285.4 points were claimed (and granted) after completion of the task. When I checked the results status page roughly 1 hour ago, the machine had reported the trickle message for 90% completion and already claimed ~380 points at that point. I'm seeing the opposite. A task that claimed 349.6 points at 90% claimed and was credited 548.4 points at completion. So there's also the possibility that prematurely terminated units will cost a lot of points. |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1405 Status: Offline Project Badges: 
|
DutchRaider wrote:
BETA_ FAHB_ avx17556-ls_ 000023_ 0002_ 001_ 0-- Jabbah Invalid 9/21/15 19:15:55 9/24/15 22:39:36 2.69 / 2.72 70.3 / 0.0 [11:25:51] INFO: received message from server to exit after next major checkpoint. [11:25:51] INFO: Checkpointed. Progress 3000 of 100000 steps complete CPU time 2909.396000 ... [14:09:05] INFO: Checkpointed. Progress 9000 of 100000 steps complete CPU time 8705.128000 [14:36:03] INFO: Exit:<current_step>10000</current_step> <total_steps>100000</total_steps> 14:36:03 (22315): called boinc_finish(0) </stderr_txt> Keith, could you have a look at DutchRaider's result? As you see, it was marked invalid, got a soft stop, stopped at 10,000 steps, but did not send a trickle message and did not uploaded the first 10% result files. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges: 
|
Hi everybody,
----------------------------------------until now, all beta WUs were valid on my side. On Linux-based host, whatever the CPU is, the duration estimation is still not accurate (effective duration is about 50-100% over the estimated time: i.e. between 25 and 30 hours) and each valid WU brings exact 422 points, regardless of the real host performance. Cheers, Yves |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1405 Status: Offline Project Badges:
|
Could somebody please kindly remind me what log flag it is used to display following information [13:29:15] INFO: Checkpointed. Progress 2000 of 100000 steps complete CPU time 861.702724 There's no log flag for that INFO. It's not in BOINC's Event Log, but finally in the Result Log here at WCG or during the run on your PC in stderr.txt in the slot directory of the corresponding task. Don't open stderr.txt during run time, but make a copy to see the contents. |
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1326 Status: Offline Project Badges: 
|
Could somebody please kindly remind me what log flag it is used to display following information [13:29:15] INFO: Checkpointed. Progress 2000 of 100000 steps complete CPU time 861.702724 There's no log flag for that INFO. It's not in BOINC's Event Log, but finally in the Result Log here at WCG or during the run on your PC in stderr.txt in the slot directory of the corresponding task. Don't open stderr.txt during run time, but make a copy to see the contents. Thanks for clearing that up for me. I returned all my work and it was between 11 and 12 hours on Windows 10 points were around 338  [Edit 2 times, last edit by Speedy51 at Sep 25, 2015 7:05:40 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Another thing I've now seen is a task terminated at 90% at 65 hours when it would easily make the 4-day deadline. Is there a way to turn off the trickle messages to avoid this?
|
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
What is that going to gain anything?
----------------------------------------1) If not completing X percent by Y time, the TIP (Task in progress) is instructed to quit at the next checkpoint, then report that through the last checkpoint. 2) Simultaneously, the next task is generated from that checkpoint, so 1) continuing beyond is duplicating effort. If you are into wasting time, then intervene and set your client to go offline, which by itself is another signal for the server as then after 3 hours/days(?) there being no contact a new task will be generated [There being a seeming goal to complete the millions of steps within 3-4-5 weeks(?). @tech: Point your ears... What I would propose to WCG is that you cap the number of tasks being assigned to a host, so it's not like one is sitting in queue for days before commenced. These will hardly get chance to get anywhere much when not started semi immediately [though you wrote the mean 100K computing time is under 24 hours]. An assignment ratio of 1:1 seem best, same as in beta... report one and give the next, e.g. after 90% is trickled. This for seamless computing on hosts that are set to only do FAHB/FAH2. [Edit 1 times, last edit by SekeRob* at Sep 25, 2015 9:35:24 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
1) If not completing X percent by Y time, the TIP (Task in progress) is instructed to quit at the next checkpoint, then report that through the last checkpoint. It seems confusing and just leads to more jobs being issued if the system sets X% > the % indicated by time Y. If Y is 3 days on a 4 day limit, it would make sense for X to be <=75. Otherwise, you get the situation I had of a WU that is 85% complete being told to exit at 90% so that someone else can do the last 10%. (And what deadline do they get? It's entirely possible that the newly issued 10% will be done later than the original machine would have done it.) |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
No, the next task does the 10k + the next 90k in the sequence for a total af 100k at again the standard deadline. Why it is confusing to you is strictly speaking not relevant. We get tasks to crunch and the techs have the mission to get the multi-million steps processed and returned in the shortest order and they do structure that based on bulk statistcs.
----------------------------------------[Edit 1 times, last edit by SekeRob* at Sep 25, 2015 4:20:39 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
No, the next task does the 10k + the next 90k in the sequence for a total af 100k at again the standard deadline. Fine. That doesn't change my point that it means unnecessarily terminating a working job that will be done within the specified time anyhow. |
||
|
|
|