21)
Message boards :
Number crunching :
Problems with Rosetta version 5.85 (or 5.86 for linux)
(Message 48876)
Posted 20 Nov 2007 by Rhiju Post: Thanks for continuing to post! |
22)
Message boards :
Number crunching :
Rosetta Application Version Release Log
(Message 48875)
Posted 20 Nov 2007 by Rhiju Post: Rosetta@home has been updated to 5.85: - New features include proper diagnostics for a different kind of symmetry ('D2') - A fix to the modeling of large symmetric complexes, and a lower memory usage by some of those jobs as well. - A new output format that may be useful for designing new proteins. - Some major improvements in the energy function used in RNA structure prediction, which we look forward to testing on a large scale! |
23)
Message boards :
Number crunching :
Problems with Rosetta version 5.81
(Message 48874)
Posted 20 Nov 2007 by Rhiju Post: We're looking into this one -- most of the failures (which are happening at a pretty low overall rate of a few percent) for BOINC_SYMM_FOLD_AND_DOCK_RELAX are indeed due to this error, which should be fixable! Thanks for continuing to post! I've had errors from hbonds.cc on some BOINC_SYMM_FOLD_AND_DOCK_RELAX-****_-crystal_foldanddock__2257 WUs: |
24)
Message boards :
Number crunching :
Problems with Rosetta version 5.81
(Message 48873)
Posted 20 Nov 2007 by Rhiju Post: Actually, I think this is our (the developer's) fault! There may be to much output from the run, and there's a maximum of something like a few Mb's allowed in the text output ("stdout.txt"). We're looking into it. Thanks much for posting -- we can't catch these kinds of problems on ralph! 120690844 |
25)
Message boards :
Number crunching :
Problems with Rosetta version 5.80
(Message 47424)
Posted 5 Oct 2007 by Rhiju Post: Hi all: We're trying to track down several sources of error. I'm not sure if anyone's posted about this, but a small number of workuntis with the batch number 2156: mcr1__BOINC_ABRELAX-mcr1_-mfr__2056_ appear to be flawed. I've cancelled the job; you should also feel free to abort these jobs if you see them. There aren't that many. I just fixed the problem and sent out a similar job with ID 2059. We're looking into a few more issues too.. I've just contacted the people in charge of the other jobs... thanks *very* much for posting! |
26)
Message boards :
Number crunching :
Problems with Rosetta version 5.80
(Message 46638)
Posted 19 Sep 2007 by Rhiju Post: Also, I think this has been posted elsewhere: if you have jobs marked CAPRI, you should feel free to abort them. Although most of these workunits are running fine and we are using the incoming data, there is a small possibility of the workunits being frozen or of the output files being so large that the file transfer can get bungled. No more of these workunits are being sent out! |
27)
Message boards :
Number crunching :
Problems with Rosetta version 5.80
(Message 46637)
Posted 19 Sep 2007 by Rhiju Post: Hi Conan, I think you're exactly right -- the WU must have frozen at some time. We previously saw a high rate of watchdog kills for this type of workunit due to freezing, so that supports your hypothesis. There's clearly some problem with these jobs -- we've discontinued them until we can fix the issue! My preferences are set to 21,000 seconds (6 hours). |
28)
Message boards :
Number crunching :
loss of credit post crash
(Message 46636)
Posted 19 Sep 2007 by Rhiju Post: Hi everybody: Thanks for your posts and for your patience over the last week. Quite a few things have been crazy. We have been testing all our workunits on the RALPH test server and they went through fine -- so your feedback over here at Rosetta@home has been critical to identifying and (in some cases) fixing new problems. The issue with the CAPRI workunits appears to be the large numebr of generated models and the size of output files; this was hammering our already frazzled fileservers. We are no longer sending out those jobs -- if we do, we'll fix this issue first. We're very sorry for this problem; it was totally unanticipated. There was also a separate issue with some workunits sent out before the crash not being accepted as valid; we had a problem with the database, and I think DK has fixed this. Then of course there was the massive outage; as BarryAZ has explained, this is causing some craziness with the credits that should hopefully be gone in a week or so. If you can, bear with us here. The results we're getting back are exciting on a number of scientific fronts. The CAPRI data on predicting protein-protein interactions is very interesting and we're analyzing it now. The work with NMR-constrained protein structural inference has the potential to revolutionize how structures are solved. And there's more exciting stuff coming soon -- we'll try to be as careful as possible! Look instead at the total work completed credits. The average is based on something like the past 2 weeks, so you would expect it to drop and 'stay dropped' until the outage timeframe begins to fall outside that two week window. Daily credits for me still haven't quite recovered to the pre-crash levels -- todays hiccups didn't help with that of course, nor did the release into the wild of some 'bad boy' work units which CPU's would chew on but not yield credit. Take a look at the message board topic regarding the 5.80 application and look thru it -- work units with 'Capri' in the title have been mentioned as work units you want to abort. |
29)
Message boards :
Number crunching :
Problems with Rosetta version 5.78
(Message 46100)
Posted 12 Sep 2007 by Rhiju Post: One more question -- did you happen to notice if the screen looked totally stuck before the crash? (Probably too much to ask.) Ok, don't want to beat a dead horse, but just noticed |
30)
Message boards :
Number crunching :
Problems with Rosetta version 5.78
(Message 46098)
Posted 12 Sep 2007 by Rhiju Post: Thanks to everyone for posting. I think I know how to fix this (the watchdog problem)! I have removed these jobs from the queue for now, and when they are sent out again, we should see fewer premature exits... |
31)
Message boards :
Number crunching :
CAPRI14?
(Message 46084)
Posted 12 Sep 2007 by Rhiju Post: This is puzzling -- those jobs arent taking long. We'll look into it. In the meanwhile, can your or other post links to the appropriate results that were killed? |
32)
Message boards :
Number crunching :
Rosetta Application Version Release Log
(Message 45713)
Posted 2 Sep 2007 by Rhiju Post: This version adjusts the way that experimental data is used to guide some RNA folding runs; there's also a new feature that lets us choose building blocks for RNA folding from different existing source structures. Also, this version should improve folding efficiency in runs that involve a special type of symmetry (''D2''). |
33)
Message boards :
Number crunching :
Problems with Rosetta version 5.78
(Message 45712)
Posted 2 Sep 2007 by Rhiju Post: Not too much different in this app from previous version. Thanks for continuing to post problems! |
34)
Message boards :
Number crunching :
Problems with Rosetta versions 5.72 and 5.73
(Message 44917)
Posted 12 Aug 2007 by Rhiju Post: Any further problems with those workunits? From our end, the overall error rate seems pretty low for those workunits. oh joy, i got about 5 of those same style of WU's on my ssytem. but its a day or so out to get to them. wonder if they are going to hang on my system or not. let you know. |
35)
Message boards :
Number crunching :
Problems with Rosetta versions 5.72 and 5.73
(Message 44916)
Posted 12 Aug 2007 by Rhiju Post: Hi -- this "warning" is OK, We'll try to remove it from later applications. 5.73 |
36)
Message boards :
Number crunching :
Rosetta Application Version Release Log
(Message 44914)
Posted 12 Aug 2007 by Rhiju Post: Rosetta@home has been updated to 5.76; as with the previous update, some older workunits will continue to crunch with 5.68. This version fixes another bug in high resolution modeling of RNA; due to a rather arcane issue, certain RNA hydrogen bonds were not being scored properly. We will also be testing a new procedure for closing chainbreaks that should lead to faster recognition that a protein structure is completely wacky, and allow your computer to go on to simulating a new conformation quickly. More efficient, hopefully! |
37)
Message boards :
Number crunching :
Problems with Rosetta version 5.76
(Message 44913)
Posted 12 Aug 2007 by Rhiju Post: Hi, this application is pretty similar to the previous one (5.73) -- thanks for continuing to post and reply to issues. |
38)
Message boards :
Number crunching :
Problems with Rosetta versions 5.72 and 5.73
(Message 44685)
Posted 4 Aug 2007 by Rhiju Post: We figured out this issue -- when this workunit gets resent, it should work fine. I contacted the person in charge of this workunit, and we won't be sending out anymore until it gets fixed... |
39)
Message boards :
Number crunching :
UPX
(Message 44684)
Posted 4 Aug 2007 by Rhiju Post: Hi -- this looks great, we'll use this compression for the next update (sorry I missed it for these last couple). It wil be nice to compress the Mac apps now! |
40)
Message boards :
Number crunching :
Problems with Rosetta versions 5.72 and 5.73
(Message 44576)
Posted 1 Aug 2007 by Rhiju Post: I contacted the person in charge of this workunit, and we won't be sending out anymore until it gets fixed... This doesn't sound right -- we're looking into this job (1850) now. |
©2024 University of Washington
https://www.bakerlab.org