|
1)
Message boards :
Number crunching :
Minirosetta 3.73-3.78
(Message 79578)
Posted 21 Feb 2016 by fractal Post: 2/20/2016 2:07:57 AM | rosetta@home | Rosetta Mini needs 57220.46 MB RAM but only 6842.83 MB is available for use. You generally need server class hardware to get more than 32 GiB of memory. <begin wry humor>And, since the project shuts you down if you fail for ANY work unit, you need 60 GiB of RAM per core. That's 240 GiB for a quad core. You can get that with AMD Opterons or Intel Xeons using registered ECC RDIM's. This is not a viable approach for most volunteers.<end wry humor> That aside, I had to manually update 8 stuck machines yesterday. I was about to say that I didn't have to restart any today but just found one on a 20 hour backoff. Fortunately I increased my buffer from a half a day to a full day to give me time to find them before they run dry. Oh, and why is it called "mini rosetta?" See https://www.rosettacommons.org/content/what-minirosetta |
|
2)
Message boards :
Number crunching :
Minirosetta 3.73-3.78
(Message 79570)
Posted 20 Feb 2016 by fractal Post: Two of my systems have started intermittently falling into 'project backoff' for 10-40 hour periods after getting this message in the logs (If I go and do a manual 'request new tasks' they successfully get more tasks but I noticed because their work queues dry out: I found two of my machines in that state this morning and several yesterday. 2/19/2016 5:54:25 PM | rosetta@home | Computation for task rb_11_07_60457_104894__t000__0_C1_beta_nov15_cart_fa_wt_0.40_SAVE_ALL_OUT_IGNORE_THE_REST_327108_852_1 finished That machine had 18 hours of backoff when I found it this morning. it still had one work unit running out of four cores. 2/20/2016 3:04:19 AM | rosetta@home | Computation for task foldit_2001101_s003_fold_and_dock_SAVE_ALL_OUT_328024_8728_0 finished This machine was completely out of work when I found it at the same time with over 24 hours of backoff. It got work as soon as I manually refreshed the project. My priority 0 backup project was not getting work either, but that never seems to work.. 2/20/2016 7:10:56 AM | Universe@Home | Sending scheduler request: To report completed tasks. I don't mind not getting a work unit that needs 60 GiB of RAM but please don't refuse to give my meager machine more bite sized work just because of that. |
©2026 University of Washington
https://www.bakerlab.org