I setup my app_config as follows, but am still only getting one CEP2 WU. Am I missing something?
Thanks,Code:<app_config> <app> <name>cep2</name> <max_concurrent>4</max_concurrent> <fraction_done_exact/> <gpu_versions> <gpu_usage>.5</gpu_usage> <cpu_usage>.4</cpu_usage> </gpu_versions> </app> <project_max_concurrent>8</project_max_concurrent> </app_config>
redmaw
go to your WCG device profiles and set project specific settings up
https://secure.worldcommunitygrid.or...viewDevices.do
I like large posteriors and I cannot prevaricate
Well I did forget about that, but it is already set to "unlimited"
Noticed the system in question is not showing up on my device list though so that is probably the real issue (account name/team are correct though). Just added it a couple hours ago so maybe it needs some more time...
Hmm just got this notice that CEP2 needs 135 MB more of disk space. Not sure what BOINC is getting at though as it still has 2.85 GB of disk space available to it.
Last edited by Red Maw; 10-19-2014 at 07:24 PM.
2.85 GB might should be enough for multiple CEP WUs... I have two on this machine using a whopping total of 386MB with all the others...
so you could have:Code:<app_config> <app> <name>cep2</name> <max_concurrent>4</max_concurrent> <fraction_done_exact/> EDIT: remember this will only take effect on cep2... you could omit <gpu_versions> there are no gpu cep2 wu's <gpu_usage>.5</gpu_usage> there are no gpu cep2 wu's <cpu_usage>.4</cpu_usage> there are no gpu cep2 wu's </gpu_versions> there are no gpu cep2 wu's </app> <project_max_concurrent>8</project_max_concurrent> only 8 cores running wcg ?? What size rig?? </app_config>
Code:<app_config> <app> <name>cep2</name> <max_concurrent>4</max_concurrent> </app> </app_config>
Rig should load a good few then, once they are at the head of the queue, process 4 at a time.
Keep an eye on it while it settles in
Adjust disk space in preferences
The list of choices is as follows:
Code:All the tags use lowercase characters. The app_name may contain uppercase characters. Note: The sections in square brackets '[foo/]' are optional. When you want to use any, remove the square brackets. Each <app> element specifies parameters for a given application: name short name of the application as found in the corresponding <name>xxxxx</name> tags in your client_state.xml file. In the above example, the application name is "uppercase". max_concurrent The maximum number of tasks of this application to run at a given time. fraction_done_exact if set, base estimates of remaining time solely on the fraction done reported by the app. New in 7.3.13 gpu_usage The number of GPU instances (possibly fractional) used by GPU versions of this app. For example, .5 means that two jobs of this application can run at once on a single GPU. cpu_usage The number of CPU instances (possibly fractional) used by GPU versions of this app. Each <app_version> element specifies parameters for a given app version; it overrides <app>. New in 7.2.39 app_name the short name of the application. plan_class the plan class of the app version. avg_ncpus the number of CPU instances (possibly fractional) used by the app version. ngpus the number of GPU instances (possibly fractional) used by the app version. cmdline command-line parameters to pass to the program. project_max_concurrent A limit on the number of running jobs for this project. New in 7.4.9 If you remove app_config.xml, or one of its entries, you must reset the project in order to restore the proper values.
Last edited by OldChap; 10-19-2014 at 10:49 PM.
My Biggest Fear Is When I die, My Wife Sells All My Stuff For What I Told Her I Paid For It.79 SB threads and 32 IB Threads across 4 rigs 111 threads Crunching!!
Getting multiple CEP2 WU's finally, still a lot of other WU's though(was hoping for just CEP2 on this system).
@OC, Unlike many around here I only have "normal" computers, with the best CPU being 4c/8t :P
Also, I was just too lazy to delete the unnecessary parts after copy/pasting lol.
Please say you're either running a RAM Drive or SSDs? (And a fast SSD at that...)Originally Posted by Red Maw
Getting multiple CEP2 WU's finally, still a lot of other WU's though
@OC, Unlike many around here I only have "normal" computers, with the best CPU being 4c/8t :P
Also, I was just too lazy to delete the unnecessary parts after copy/pasting lol.
Otherwise the IO will gum up the system and even though you'll be fully loaded, you won't get 8 days/day you run it...
Yes I am trying out a software ram drive. Most of the CEP2 WU's (and one FAHV) are ending up as error or invalid though so far :\
Aww crap... hopefully someone around here could help you get it fixed? Cuz I'd be very interested in hearing the results...
(Start a thread about it - get some help debugging - then show us the numbers!)
do you have proper read/write access for your ramdisk?
here's how almost all my windows rigs are set up
I like large posteriors and I cannot prevaricate
How does that work for you so far? Any what happens in the case of power failure? Is there any way to back it up every hour or so?
it works perfectly
you pretty much lose all the work it was working on
you could use windows scheduler tool for it
I like large posteriors and I cannot prevaricate
Hmm... seems like we could write a script to set all of that up automatically... and add the auto hourly backup! Maybe over christmas break...it works perfectly
you pretty much lose all the work it was working on
you could use windows scheduler tool for it
Thanks for the info!
Not sure how long result status keeps data but right now it is 8 error, 1 invalid, and 10 valid for CEP2. Going to monitor this a bit more before getting concerned; I've noticed this system has a few more errors and invalid then my older one.
I am using the softperfect ramdrive since in my tests a couple years ago it had the best performance. It also allows you to save the "disk" to an image at a user specified interval.
Sounds good - If we can help, post up! I want to see some numbers with this as I'd love to do more CEP2 WUs...
Sadly the servers are not really giving me more CEP2 wu's. Sometimes I get 4 at once, but no more until those are finished, and other times I get none at all. Maybe it's because I have too many errors and invalid (up to 9 error and 4 invalid now).
Are you running "other projects when none available?"Sadly the servers are not really giving me more CEP2 wu's. Sometimes I get 4 at once, but no more until those are finished, and other times I get none at all. Maybe it's because I have too many errors and invalid (up to 9 error and 4 invalid now).
Help get your valid count up... And actually see if its erroring out on all of them or just CEP2.
Also, are you doing this on multiple machines or just one? Are they all erroring out the same way? Could it be a bad install?
This machine has 2 non-CEP2 errors and one non-CEP2 invalid. I am running other projects since I checked "run other projects when none available" so I have quite a few non-CEP2 valid WU's.
My other system has one invalid and one error (neither CEP2). Surprised to that too since I have never seen it return an error or invalid before (maybe I just didn't check often enough though).
How can you check the error? I just look at result status and it doesn't say what the problem was.
CEP2 WUs can require up to 2GB of RAM per thread. If you are using a software RAMdisk and don't have a boatload of RAM (as well as plenty of RAMdisk space to store the CEP2 data) then they fail. I actually just explained this in another thread 2 days ago.
I have plenty of ram on this system (by my standards at least). The ram drive is 4 GB and still has 3.5 GB free and the system still has 6.5 GB of free ram when crunching.
I realized last night that I am only getting 4 CEP2 WU's at once because my app_config sets the max concurrent CEP2 to 4. Took that out just now so we'll see in a few hours if I stop getting other projects.
When I was experimenting with my RAMdisk that was an actual SATA drive the calculations I came up with for my 4c/8t machine to be able to do all CEP2 from RAM was going to be in excess of 32GB of RAM. Of course, that instantly puts you into the high-end (read: expensive) CPUs and RAM sticks, which makes it a non-starter too. To do less meant CEP2 work units would terminate and you'd get no credit. For me, if memory serves me right, the WUs will slowly take more and more RAM. Do note that RAm that is listed as "free" is in indicator, but they can still terminate with several GB of RAM available depending on your settings and such.
Definitely don't let me discourage you. Things cound have changed for CEP2 WUs in the last couple of years and I may be in error too. I just remember my system with 4c/8t and 32GB of RAM wouldn't cut it.
Look at how much RAM you allow BOINC to use both while computer is in use and when idle... that could be causing it...When I was experimenting with my RAMdisk that was an actual SATA drive the calculations I came up with for my 4c/8t machine to be able to do all CEP2 from RAM was going to be in excess of 32GB of RAM. Of course, that instantly puts you into the high-end (read: expensive) CPUs and RAM sticks, which makes it a non-starter too. To do less meant CEP2 work units would terminate and you'd get no credit. For me, if memory serves me right, the WUs will slowly take more and more RAM. Do note that RAm that is listed as "free" is in indicator, but they can still terminate with several GB of RAM available depending on your settings and such.
Definitely don't let me discourage you. Things cound have changed for CEP2 WUs in the last couple of years and I may be in error too. I just remember my system with 4c/8t and 32GB of RAM wouldn't cut it.
Yeah, I tried to maximize what was available for WUs. I didn't even have monitors or keyboards attached to the machines so I set everything to maximize WU processing.
Posting Permissions
Reply With Quote
Bookmarks