Ok, just got an email from the folks at Harvard that they believe they may have found the issue from one of their upload servers. Please let me know if you are able to upload now.

Thanks,
-Uplinger

Is there any reason why I only get 1 WU in cache since the server move?

My settings haven't changed and should have 6 downloaded at anyone time.

regards,
Patrick

What prints in the event log when you only select cep2 in the relevant profile and then hit project-update in the agent? There's so many reasons why not receiving what you expect to include overworked wcg as project, buffer full due to over-projected run-times to name but a few. Have 2, want 2 over here and that's only because it otherwise will not send any due low bandwidth perception of the agent.

But, has the weight been changed, or was depletion of cep2 and then clients trying to start 4-8-16 concurrent driving these nodes nuts in disk i/o competition, the stats substantiate something is amiss? Before the server move 26-30 y/d, now barely 14-19, actually slight going down again day on day since the 9th.

Date	        CPU Years  	Results
20140428	28:355:13:21:09	44916
20140429	30:355:15:28:11	46480
20140430	26:288:23:19:36	39136
20140501	27:308:14:36:12	40754
20140502	28:138:08:26:58	40598
20140503	23:016:20:36:27	32068
20140504	13:110:05:42:05	18249
20140505	7:292:00:03:48	11002
20140506	1:002:00:00:23	1380
20140507	14:207:22:45:35	20990
20140508	18:064:01:29:44	26226
20140509	19:273:11:21:15	27577
20140510	18:183:06:31:07	28164
20140511	16:242:01:50:28	32263
20140512	15:156:23:11:58	30554
20140513	14:256:13:39:16	31018

Whatever the reason some are not getting what they want, increased one profile from the previous post noted 4 to 5 now because the run-times are short so on a 24 hour buffer there's no back to back processing of one cep2, hard-restricted in app_config. The fifth came fairly promptly. It's not the job runtime though, there's a clear per day reduced results count in line with the years processed.

Talk to us, please.

Si signor, the weight up has worked to boost the daily from 14.4 to 18.0 the successive day, but, the runtimes are so short that getting it to 25 years probably needs the allocated slots to be increased further. Had to set the device profile to 9 cep2 in order to push a back to back processing of these tasks on 1 core, relying on the fifo method to ensure the cep2 tasks are virtually always oldest in the buffer which is set to hold 24 hours worth of work for all threads.