|
21)
Message boards :
Questions and problems :
Task exited with zero status but no 'finished' file... (almost) all of them
(Message 47349)
Posted 17 Jan 2013 by kdsjsdj Post: @kdsjsdj: Noooo. |
|
22)
Message boards :
Questions and problems :
Task exited with zero status but no 'finished' file... (almost) all of them
(Message 47343)
Posted 17 Jan 2013 by kdsjsdj Post: I thought it's the clock going forward that does that? I think when BilBG said "if the clock goes backwards by more than 30 seconds" he meant "if the clock falls behind by more than 30 seconds"? Dunno, I don't want to put words in his mouth :-) Anyway, if there are methods to track time independently of the system clock (and of course there are--the High Precision Event Timer is just the latest in a whole series of kernel timers, old and new, but all with millisecond precision or much better) wouldn't it make more sense to track time using one of these instead of the system clock for something as critical as the continued survival of a science app? That would work on Windows but not Linux or OSX so they won't do that. Remember, it has to be cross-platform compatible. And if these exit with zero errors occur at the drop of a hat, shouldn't there be a way to ignore them completely, or at least remove the 100 error limit for long-running apps? Why? It works perfectly for everybody except you. If you would fix the cause of the problem then you won't need the exception you ask for. |
|
23)
Message boards :
GPUs :
Intel GPUs
(Message 47330)
Posted 17 Jan 2013 by kdsjsdj Post: Really?!! Stupid me I just assumed they would have OpenCL. Well then it's hopeless. Ubiquitous as they are or may become they'll never crunch a BOINC task. |
|
24)
Message boards :
Questions and problems :
Multi-core Computing
(Message 47326)
Posted 16 Jan 2013 by kdsjsdj Post: Can't the BOINC project kind of thumbscrew its most popular projects into doing this? How? And why? Some algorithms simply do not lend themselves to multi-threading so there is absolutely 0 benefit to be gained from making them mt. You want BOINC to thumbscrew projects into spending many man hours on converting their app to mt when it won't help them at all? Nope! That'll never happen. |
|
25)
Message boards :
Questions and problems :
Task exited with zero status but no 'finished' file... (almost) all of them
(Message 47324)
Posted 16 Jan 2013 by kdsjsdj Post: WTF? +1 |
|
26)
Message boards :
Questions and problems :
Task exited with zero status but no 'finished' file... (almost) all of them
(Message 47323)
Posted 16 Jan 2013 by kdsjsdj Post: Well, today was defrag day, and defrag hour passed by without incident. You were lucky that time. Next time you might not be. Another few things I'm wondering about: I'm a Linux user so not fully aware of all the implications of "Realtime priority application" but I would say there is a good chance it's blocking BOINC. You have the ASIO app, the chances of anybody else here having it is slim so YOU are the one who needs to test and verify. If the problems occur only when ASIO runs then you've found the problem. In that case you can specify ASIO as an exclusive app in cc_config.xml which will cause BOINC to automatically suspend whenever ASIO runs and resume whenever ASIO is not running. I'd closed and restarted BOINC to do some of my own troubleshooting and now I can't see the old log entries. I know they're kept in a file somewhere but what's it called? .../BOINC/stdoutdae.txt 2. One of the lines in the stderr for the failed CPDN task mentioned not being able to get a lock on a file. I went to my ProgramData directory and noted in its Properties that Read-Only was marked with a square, which if I understand correctly means that some items inside had been marked read-only... so I cleared the checkbox, told windows to make the change (to writeable) to all files in the directory and subdirectories... but when I opened its properties again, the checkbox was marked with a square again... That kind of crap rarely happens on Linux (hint, hint) While I'm at it, let's mention a little fiasco I had a while ago: If you're worried about it then wipe the slate clean and restart: uninstall BOINC, delete ../../BOINC/, reinstall 3. Speaking of permissions Mess around with that stuff and you're on your own. Nobody here wants to second-guess what you did or did not do and what effect that might or might not have. The permissions work. Don't mess with them. Same with client_state.xml... best just leave it alone unless you really know what you're doing. |
|
27)
Message boards :
GPUs :
Intel GPUs
(Message 47322)
Posted 16 Jan 2013 by kdsjsdj Post: what would one bring as sales pitch to get them to port to the SB/IB GPU? Easy... ubiquity. |
|
28)
Message boards :
Questions and problems :
Task exited with zero status but no 'finished' file... (almost) all of them
(Message 47310)
Posted 16 Jan 2013 by kdsjsdj Post: Doh! Right, user mode. The net command won't work, use boinccmd.exe instead for stopping and then invoke boinc.exe to restart. If your binaries are in c:\program files\boinc' then use "c:\program files\boinc\boinccmd" --passwd <password> --quit to stop the client and "c:\program files\boinc" --detach to start it. <password> is found in the file gui_rpc_auth.cfg in the BOINC data dir which is c:\programdata\boinc\ by default. |
|
29)
Message boards :
Questions and problems :
Task exited with zero status but no 'finished' file... (almost) all of them
(Message 47307)
Posted 16 Jan 2013 by kdsjsdj Post: ... these last 2 suggestions sound like extreme measures to me. Defrag is scheduled to run once a week; how do I tell BOINC to stop running while this is happening? A weekly defrag is more than most people need but if that's what you want then go for it. It sounds like you're running the defrag via Event Scheduler, if so then I would create a .bat file that: 1) shuts BOINC down 2) runs the defrag 3) runs a virus scan 4) other misc. weekly maintenance 5) re-starts BOINC Then run that .bat in Event Scheduler. The BOINC service can be stopped with the net command, not sure if it's "net BOINC stop" or "net stop BOINC" but you'll figure it out. And it might be boinc-client instead of BOINC. You can start the BOINC service with "net start". Again you'll have to experiment to find the correct syntax. And are you telling me that I can't even have the luxury of knowing that my computer clock is accurate if I want to run BOINC reliably? Care to explain how time synchronization relates to this problem? If your clock lags behind a fair bit and many do, we've found this out through experience, then when it syncs it makes the science apps think they haven't received a heartbeat from client for more than whatever the limit is. In fact the limit hasn't been exceeded it's just that the clock jumped ahead a large amount. See what I mean? This is a well known cause, confirmed and verified many a time. Some people with clocks that lag a lot find that slipping a new battery in cures the problem. Others find that resyncing just once a day instead of evry 5 minutes or so cures the problem. Yah, they still get a "exit with finished file" but it happens only once per day instead of every 5 minutes. Anyway, it's been over 24 hours and I haven't had an exit with zero error since the CPDN WU died. Well then it seems to me the CPDN wu was the cause or somehow connected to the cause. I don't know what I changed that solved the problem; I didn't act on (1) and (4) until a few hours ago, and BOINC had been running flawlessly for several hours by then; perhaps CPDN was the cause of the errors, and its WU crashing out "solved" the problem; or perhaps it was this discovery that fixed things: before, when I set the GPU to "run always", the computer would be very unresponsive when I'm at the keyboard, to the extent that sometimes I couldn't even open BOINC manager to stop the GPU for want of the GUI responding for whole minutes at a time. The last error occurred about the time when I applied the TThrottle "fix". I can't be sure whether it was before or after the fix though. hehe you've got your work cut out for you. I can't put my finger on the problem for you but I can tell you some methods that work for me when things get as complicated as your situation: 1) assume NOTHING, don't say "it can't be that" until you've tested and verified 2) reduce complexity by running only 1 project for a few days or even a week, if everything goes well then add another project, run the 2 for a week, add another and so on... if you add project X back into the mix and that's when your troubles reappear then it's likely X or something related to X so then set your other projects to No New Tasks and run just X, maybe start with just 1 X task, then 2, then 3... |
|
30)
Message boards :
Questions and problems :
Task not paused when computer in use
(Message 47306)
Posted 16 Jan 2013 by kdsjsdj Post: I recently noticed the following logging options for cc_config.xml: <app_msg_receive> (Shared-memory messages received from applications.) <app_msg_send> (Shared-memory messages sent to applications.) If you turn them on you might be able to see whether the client is or is not sending a suspend order and whether the application acknowledges receiving it or not, etc. I don't know the details of the protocol between client and apps but someone around here probably will. |
|
31)
Message boards :
The Lounge :
The Seti is Slumbering Cafe
(Message 47295)
Posted 15 Jan 2013 by kdsjsdj Post: And BTW, calling me a young man would be like calling George W Bush "intellectual". ROFLMAO! OK, we'll get along fine. |
|
32)
Message boards :
Questions and problems :
Task exited with zero status but no 'finished' file... (almost) all of them
(Message 47294)
Posted 15 Jan 2013 by kdsjsdj Post: I suspect what is happening here is that BOINC is too busy with itself running time estimations to notice the WUs prodding it for attention. That proves you grasp the concept but when I look at the times in the log snippets you posted all the estimation operations are stamped with 8:43:21 so it appears the time estimations seem to take only 1 second or less. I'm not 100% sure but I think the apps are not that impatient. IIRC, they check for a heartbeat from BOINC client every 60 seconds. |
|
33)
Message boards :
The Lounge :
The Seti is Slumbering Cafe
(Message 47273)
Posted 15 Jan 2013 by kdsjsdj Post: Oh no! They're taking boots and jackets off. Hey you guys! We have no food, our women are old and our wells are all dry. They have food, good ale and fast women in the next town only 40 km west. And someone spotted an LGM there last week. Really! |
|
34)
Message boards :
GPUs :
TThrottle as transparent GPU computing solution!
(Message 47267)
Posted 15 Jan 2013 by kdsjsdj Post: I wonder if Fred can talk to the BOINC team and show them how he does this GPU throttling trick so they can implement this transparent GPU computing that's been on their wishlist... They already know how to do it. |
|
35)
Message boards :
GPUs :
HD5450 on Fedora 17 - No usable GPU detected
(Message 47266)
Posted 15 Jan 2013 by kdsjsdj Post: Even the easiest option is to uninstall this BOINC and install Berkeley BOINC instead. That one will be able to detect the drivers, as it runs under your user account. That rarely works. The BOINC installed by the Berkeley installer is built for Ubuntu. He's running Fedora. Fedora won't have the same shared libraries (dlls) as Ubuntu therefore the Berkeley BOINC won't run on Fedora 99% of the time. There are 3 ways around this problem: 1) Do NOT uninstall the BOINC installed from repos, simply disable the BOINC daemon so it does not start at boot time. Then copy the binaries (boinc, boinccmd and boincmgr) from /usr/bin/ to ~/boinc/ and run them from that directory. ~ is shorthand for your home directory. If you start the client then it runs on your account and in a directory you own and there are no permissions problem. Basically what you're doing is installing binaries from the repos in a Berkeley fashion. The repo installer also installs all the shared libs required for the repo binaries so no problem there either. You can make the client and/or the manager autostart at bootime or you can start them manually. However if you uninstall the repo installation then you also uninstall the shared libs BOINC needs so don't uninstall 2) Give the boinc user (the user created by the package installer) permission to access the video subsystems. I don't remember how to do that for Fedora, I had it all documented, tested and confirmed once and was ready to add it to the wiki then other crap happened and I forgot and eventually lost the files and links and so on. IIRC, Dagorath posted the method here once, could do a forum search I suppose. People at GPUgrid would know how too. 3) Download the BOINC source code, compile it and install it Berkeley fashion on your account as described in 1) above. You'll likely have to dig up and install some shared libs but lots of people do it. If you want to run BOINC on Linux then the easiest and best way is to install Ubuntu LTS. The Berkeley BOINC builds are guaranteed to run on Ubuntu LTS and if you want a daemon BOINC install from repos it seems like the Ubuntu BOINC package installers do the best job of keeping up with BOINC updates. |
|
36)
Message boards :
BOINC Manager :
System indefinitely slow after deactivating GPU
(Message 47263)
Posted 15 Jan 2013 by kdsjsdj Post: They have OpenCL but it doesn't seem to work worth a damn on their GPUs from what I hear. |
|
37)
Message boards :
The Lounge :
Gerry Anderson, Thunderbirds creator dies, age 83
(Message 47262)
Posted 15 Jan 2013 by kdsjsdj Post: Oh wow! He created Supercar? Mike Mercury, Dr. Beaker, Dr. Popkiss and Jimmy. Oh and Clear-VU!!! About 5 years before Supercar we didn't even have TV in the city of 30,000 where I lived. |
|
38)
Message boards :
The Lounge :
The Seti is Slumbering Cafe
(Message 47260)
Posted 15 Jan 2013 by kdsjsdj Post: Oh LGM Where art thou? SETI can't find thee They know not how. |
|
39)
Message boards :
BOINC Manager :
System indefinitely slow after deactivating GPU
(Message 47255)
Posted 14 Jan 2013 by kdsjsdj Post: Problem with HCC-GPU tasks, which use OpenCL, is that NVidia cards are just not very good at crunching these and using the PC at same time. Because nVidia is stuck on CUDA and hasn't embraced OpenCL. I used to be an nVidia fan but if they don't get with OpenCL soon I'll buy AMD-ATI instead. |
|
40)
Message boards :
BOINC client :
Anyone running ubuntu in 4gb flash with latest boinc & cuda?
(Message 47254)
Posted 14 Jan 2013 by kdsjsdj Post: Hopefully apt-get is included in the min imstall. I can't imagine it not being there but if it isn't then remember apt-get is just a frontend for synaptic which will most definitely be there. It's not as easy as apt-get but you could use it to install apt-get... |
Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.