I looked at a few I had done and I had one that everyone(me + 4 others) had errored out on it. Then I had one that the four pevious people had not had enough memory to run the WU!Originally Posted by STE\/E
I think I have had about 3 in 1200+ that errored out so far.
BOINC Sees it - BOINC DOES it!
I figured out what happened, some Lattice Garlic Wu's decided to start up & were sucking up their usual 9-10gb of Memory, so with that & the 1gb+ memory each NFS were sucking up on 8 Threads each was to much memory usage & the Wu's started erring on those Box's. I had to trash the Lattice Wu's & now the NFS Wu are running okay again ...
Wow!!! 9-10GB of RAM for those suckers? Must be some serious data in those WUs or something... glad you figured it out though and can claim your rightful place atop the challenge list. Well, first after Mumps that is!. It's kinda like seeing the afterburners light up on Top Gun when he feels like turning them on!
I crunch as if there is always an NFS challenge.
Yep, 32GB for 24 cores. The wu error right at the time the progress meter usually moves off 0% so I am only loosing 3-4 min per task. I'm going to tear the thing apart tomorrow and reinstall all of the memory. Got some new memory on the way also.
Just to let you guys know that Greg is committed to free up some space on the server to make it more stable. I already downloaded and backup all external work that I need to run for NFS@Home (post-processing work). That's about 30-40 GB data he can spare to the challenges.
Thanks, that is a big help. I know some of the guys with bigger farms are pretty reluctant to crunch a project when it's less than 100% because of the backlog it does to their rigs!
I did my part and backups. I am just waiting for one of his replies to me to see what can we do to make the server even more stable.Thanks, that is a big help. I know some of the guys with bigger farms are pretty reluctant to crunch a project when it's less than 100% because of the backlog it does to their rigs!
There's a lot going on on the back stage of NFS@Home that I try to post here so that people can understand what's going on.
Greg, NFS manager, and post-processing helpers (like me), are trying hard to watch for a healthy server. Bandwidth slot for us are being shrink so that download and upload of wu's are optimal faster for you guys. I also warned my mates to backup the data files so Greg can free up more and more space to the server. Every little help is appreciated.
Anyway, from SETI.USA side we are doing a great job. I just regret I can't help on running BOINC but I am helping on post-processing stage.
Carlos
I'll probably have my 5 Linux Box's on the Project until I can get all the APP's over 5000 Hr's, once that happens then I'll be done with the Project possibly forever ... Way to much demand on the systems with the Memory usage & the usual amount of error's ...
The linux version is faster because it is 64 bit compatible. As I said previously a new win64bits build was achieved but it needs testing.
About the errors, when sieving you guys need to reduce the system overclock because the client is very sensible. It is almost as sensible as running an LLR test (prime hunt).
I never had errors while running NFS because my system is stable, I have plenty of memory and I only attach to BOINC one project at the time. I don't share BOINC with a bunch of projects. Maybe there are conflicts between projects.
Carlos
Posting Permissions
Reply With Quote
