|
41)
Message boards :
Questions and problems :
Boinc breaking down / reinstall needed
(Message 41896)
Posted 2 Jan 2012 by (banished: ID 2516) Post: The first thing I would try is an update to the recommended BOINC version. |
|
42)
Message boards :
GPUs :
Others running HPC type systems CXT8000/VDACr8
(Message 41894)
Posted 2 Jan 2012 by (banished: ID 2516) Post: Well, when you find a link to the 16 slot extender let us know. As for using Teslas, I don't think many volunteers are using them for existing BOINC GPU projects because none of the existing GPU projects has a science app that can take full advantage of a Tesla so they end up performing about as well as a high end GTX. That's just what I heard and it might not be true. |
|
43)
Message boards :
Questions and problems :
running 2 BOINCs ?
(Message 41889)
Posted 2 Jan 2012 by (banished: ID 2516) Post: Man you are nuts but that doesn't mean your idea won't work ;-) Open a terminal (command prompt), cd to the appropriate directory (probably c:\program files\BOINC) and run "boinc.exe --help" and "boincmgr.exe --help" and note that the client and the manager both have options that allow more than 1 instance of each to run simultaneously. But there is more to it than that, you have to have each client-manager pair communicating on its own TCP port and using its own data directory. I believe all the command line options are there to allow you to do exactly what you want except for the CPU affinity business. If you can't figure it out from there I can provide some example command lines. I don't know if you can set processor affinity on Windows the way you want but I could be wrong (I do Linux not Windows). |
|
44)
Message boards :
GPUs :
Others running HPC type systems CXT8000/VDACr8
(Message 41888)
Posted 2 Jan 2012 by (banished: ID 2516) Post: I like the attachment unit that uses a PCIe extender card to add 16GPU's to an existing system, I could use this on a Mac Pro and have 16GPU's on top for an existing 24 core Xeon. You're going to have 16 high end GPUs plugged into 1 PCIe slot? I don't think 1 slot can provide the bandwidth 16 fast GPUs require. |
|
45)
Message boards :
Questions and problems :
running 2 BOINCs ?
(Message 41886)
Posted 2 Jan 2012 by (banished: ID 2516) Post: Which OS are you using? Are you able to use a command line? Do you want the 2 independent clients to run simultaneously or just one at a time? Why do you want 2 independent BOINC clients? Maybe there is a simpler way. |
|
46)
Message boards :
Questions and problems :
GPU
(Message 41883)
Posted 2 Jan 2012 by (banished: ID 2516) Post: Depends which OS you're using. I installed Kubuntu 11.10 (Linux) recently and my NVIDIA GPU was automatically setup for crunching when I installed the OS. All I had to do was attach to a GPU project that supports NVIDIA cards and it started crunching. I don't have an AMD GPU so I don't know if it would have done the same for AMD but probably it does. If you're using Windows then you have to install the proper driver for your GPU manually. For NVIDIA you need to install their CUDA driver and for AMD I believe it's their OpenGL driver. The drivers are available at the respective websites. For NVIDIA one driver works for all their CUDA capable cards which means nearly all their cards. I believe they have a list at the website to tell you which cards are CUDA capable. For AMD I think there are different drivers for different cards but I'm not sure. Don't know anything about OS X. A recent version of BOINC is recommended, preferably the latest recommended version. In your BOINC preferences you can select whether to use the GPU or not so check those preferences and set accordingly. |
|
47)
Message boards :
Web interfaces :
[wish] Allow users to temporarily turn on resend lost work units or to declare failed computers in their account
(Message 41866)
Posted 1 Jan 2012 by (banished: ID 2516) Post: You didn't have to do any of that. You could have just aborted the tasks you restored from backup. Eventually the server will send the tasks to another computer and life will go on. |
|
48)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41842)
Posted 28 Dec 2011 by (banished: ID 2516) Post: Congrats dwarthog!! That's the way ya do it. I said I would explain why you should not do the script thing so I will. Starting processes as service daemons is traditionally done using a Sys V protocol/standard which requires a startup script that meets certain specifications. Your startup script does not meet those specs so I recommended not doing it just in case it messes something else up. Sorry if my first reaction was a little too strong/abrupt but I was very tired. Now that I think about it, it probably would just fail and not do anything but I was thinking it might render the system unbootable. |
|
49)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41833)
Posted 28 Dec 2011 by (banished: ID 2516) Post: Noooooooooo! Don't do that script thing!!! I'm too tired to explain now but I will tomorrow. Don't mess with anything in the daemon installation and don't try to run anything to do with the /var/lib/boinc-client dir. I may have a way to get that working but I'm too tired to explain now. |
|
50)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41830)
Posted 28 Dec 2011 by (banished: ID 2516) Post: The experiment in /clienttest looks good and I kind of think it was just a wu/app problem or maybe because you didn't have the 32 bit compatability libs installed for climate prediction. I would do ls -l /var/lib/boinc so you can see all the links in that dir and the files they point to. I would delete all the files the links in /var/lib/boinc point to then everything in /var/lib/boinc itself including subdirectories. Then set the service to start at boot time, reboot and try ps aux | grep boinc and/or start the manager. If you're going to add climate prediction again install the 32 bit compatibility libs first, the names are here but I think the names of the packages might be out of date. It might be a good idea to test the daemon start with some project other than climate prediction. |
|
51)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41819)
Posted 27 Dec 2011 by (banished: ID 2516) Post: It's pretty obvious you've got another instance of BOINC running and Ithink the reason is that you haven't stopped BOINC from starting as a daemon when you boot. Let's make sure that it won't start again when you boot. Do sudo /sbin/chkconfig boinc-client off and post the output from that command here, please. Then restart the OS. You have to restart it and not just log out and log back in because the daemon will continue to tun even if nobody is logged in. After you restart do ps aux | grep boinc and make sure there are no boinc processes running. If there is then you didn't do the chkconfig command properly or you didn't do a restart or something is really fubar about your system. Please do the above and nothing more. Let's see what happens with that and decide what to do from there. I think there is something screwed up with your climate prediction task(s) and I think you're going to have to kiss them goodbye. Out of all the BOINC projects that is the hardest one to run. If you just look the wrong way at one of their tasks they crash and screw up all sorts of stuff. When you have one of their tasks in progress don't crunch any other tasks along with it. If you have to stop BOINC to do a BOINC backup then make damn sure you don't stop it anywhere near the trickle up date, otherwise don't pause it, snooze it, hibernate it or turn off your computer. Don't change anything or update BOINC or update the OS. Don't even change your socks or your girlfriend because if you do you'll have no end of trouble from that task. Wait until the task is done, uploaded and reported and then change whatever you need to change. |
|
52)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41813)
Posted 27 Dec 2011 by (banished: ID 2516) Post: The event log shows it ran to the "Not using a proxy" line. After that it should restart the tasks and that's where it fails with a segmentation violation which usually means something is trying to access memory outside of its allotted memory. I don't have a clue how to interpret the stack trace messages but Trog Dog or someone else might, or you can post them in a Linux forum. You could try deleting all the files out of ~/BOINC/ and then start the client with ./boinc again and see if it crashes in the same place. I am thinking maybe it's crashing due to a problem starting the climate prediction task so remove that and everything else except the 4 BOINC binaries and see if it still crashes. If it doesn't crash then as I recall, the climate prediction app is a 32 bit app so you need to install the 32 bit compatibility libraries to run it on your 64 bit system. Not having those 32 bit compatibility libraries shouldn't crash BOINC client but ya never know. |
|
53)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41810)
Posted 27 Dec 2011 by (banished: ID 2516) Post: Thanks for the suggestion Trog Dog, but I am not convinced that it is the problem here. These files/folders are all listed as executable by others, even though the owner is root. Also, when I try to execute them as regular user, the error message I am getting is not a permission issue, it specifically reports "Another instance of BOINC is running", although ps -aux |grep boinc (or BOINC) reports none. I still tried chown'ing them to user boinc, and it made no difference in the behaviour BTW. The boinc-client, boinccmd and boincmgr binaries were installed by the package manager which runs on the root account so they are owned by root but anybody, including the user named boinc, can execute them. The files boinc-client creates will be owned by the account that runs boinc-client. When you issued the ./boinc --dir /var/lib/boinc command the client_state.xml file and other files would have been created with you as the owner. If the files had been created when boinc-client started as a daemon at boot time then boinc would own them. However, we can see that boinc-client doesn't run long enough when it's started as a daemon to write those files so we know boinc doesn't own them. I'm not sure why you get that ghost/defunct boinc process when you have it start as a daemon. I've never seen that before and you may need to find someone with more knowledge of Linux than I have to help you figure that out. Since you can start boinc manually from the command line it seems there should be no reason why it can't start as a daemon either. But at least we know you have all the shared libs required and they aren't corrupted. If you are intent on having it start as a daemon then I would recommend uninstalling it with "yum uninstall boinc" or something similar then search your disk as root for boinc* and delete any BOINC related files you find. I would do the search as root because you may not have permission to read the directories where those files might be located. I think "sudo find / boinc*" would do the job. Then reinstall the package and try it again. If reinstalling the package doesn't work and you still get that ghost process then the only other solution I can offer you is to set it up like a standard BOINC installation as installed by the Berkeley installer. Basically you would create a directory named BOINC in your home directory, drag copies of the 3 BOINC binaries from /usr/bin to the BOINC directory and make sure you own those binaries. Then copy all the files and subdirectories from /var/lib/boinc-client into BOINC and change ownership of them to you with "chown -R hakova:hakova *". At that point you can start boinc manually from the command line with ./boinc --daemon ( the daemon switch doesn't run it as a daemon, it merely runs it in the background) and then start the manager manually with ./boincmgr. If you want you can put ./boinc --daemon in a startup script to have it start at boot time and put a link to the manager on your desktop so all you do is click the link icon to start the manager. Because the installation from repos is left intact you would still get BOINC updates from repos if there should ever be a BOINC update for Fedora 16. It seems like the package managers never bother to issue BOINC updates until the distro issues an update. |
|
54)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41809)
Posted 27 Dec 2011 by (banished: ID 2516) Post: is there a way to use my old /var/lib/boinc directory, so I continue my previous tasks? That diractory is renamed to boink now in same location. That would get the client crunching his old tasks/projects but I believe he wants to have the client start as a daemon at boot time rather than starting it manually every time he boots. The daemon script at /etc/init.d/boinc-client which starts the boinc daemon issues the boinc command with --dir /var/lib/boinc rather than /var/lib/boink. If he changed the directory in the daemon script (plus one other location I can't remember) to /var/lib/boink then it would work. But it's best not to mess with that daemon script. |
|
55)
Message boards :
BOINC Manager :
(debian) not connecting to client
(Message 41808)
Posted 27 Dec 2011 by (banished: ID 2516) Post: I never would have thought of that one but it makes perfect sense. |
|
56)
Message boards :
Questions and problems :
Debian: Error -155: authentication error
(Message 41807)
Posted 27 Dec 2011 by (banished: ID 2516) Post: The authentication error is because boinccmd cannot find the password file which is named gui_rpc_auth.cfg. Some boinccmd options require the password, some do not, --project_attach is one that does. Whatever password is in that file is the password BOINC client expects to receive from boinccmd and BOINC manager. If it doesn't get that password then the client will not allow the connection. If you installed BOINC from the Berkeley installer (the .sh installer) then boinc (the client), boinccmd, boincmgr and that password file are all in the same directory so all you need to do is cd to that directory before you run boinccmd. That way boinccmd will find the password file, read the password from it and use that password to connect to boinc. If you installed BOINC from Debian repositories then one of the following suggestions may help. 1) put a link to /var/lib/boinc-client/gui_rpc_auth.cfg in your home directory, I think boinccmd will look in your home directory for the password file, make sure the password file has read permission set to all (sudo chmod a+r /var/lib/boinc-client/qui_rpc_auth.cfg) 2) cd to /var/lib/boin-client before running boinccmd 3) stop BOINC client, use a text editor to change the password in gui_rpc_auth.cfg to something you can remember, give that password on the command line like ./boinccmd --passwd {password} --get_state |
|
57)
Message boards :
Questions and problems :
Mac + BOINC as a daemon + VirtualBox
(Message 41806)
Posted 26 Dec 2011 by (banished: ID 2516) Post: If you don't want to run the "enabling script" (and I can see why you may not be able to do that) then the only way for you to crunch T4T tasks is to install VirtualBox, create a VM (virtual machine) of your own, install OS X, Linux, or Windows in that VM then install BOINC not as a daemon in that VM. You'll have a VM running inside a VM but it works. There is considerable overhead required for the nested VMs but if you have a fervent desire to crunch T4T then it's your only option other than using the enabling script. Instructions for creating your own VM are found on the T4T website forum in a thread named "How to run more than one T4T task". If you want to install Linux in the VM you create then install 32 bit Ubuntu 10.04 for best performance and a problem free BOINC install. |
|
58)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41788)
Posted 26 Dec 2011 by (banished: ID 2516) Post: Mine also says linux-vdso.so.1 => (0x00007fff0eeac000). If a library is missing it will say like: lib-xyz.so.1 => not found So you have all the libs, that's one cause eliminated. What I would do next is 1) /sbin/chkconfig boinc-client off (to prevent starting boinc at boot time) 2) reboot the OS 3) delete the lockfile in /var/lib/boinc 4) cd to /usr/bin/boinc and do ./boinc-client --dir /var/lib/boinc You might need to be root to delete the lockfile and run boinc-client. I think boinc-client might be a link that points to boinc. May sure the link still points to boinc and that root has execution privileges on boinc. |
|
59)
Message boards :
BOINC client :
Recent boinc client problem in Fedora 16
(Message 41785)
Posted 26 Dec 2011 by (banished: ID 2516) Post: The first line of output from the ps command looks right and indicates the client is running. The second line (which says defunct) indicates the client is not running. I've never seen that before so I am not sure exactly what it means. The fact that /var/lib/boinc has only the lockfile indicates boinc started and ran long enough to write the lockfile then died. It should also write two log files named stdoutdae.txt and stderrdae.txt but I'm not sure if it writes those in /var/lib/boinc or some other directory. Please search your disk for those two files and see if they have any clues. Your system logs might have some clues too. You could also open a terminal, cd to /usr/bin/boinc and attempt to run the boinc binary with ./boinc. It's output/errors will be written to the screen and will probably give some clues. You can also do ldd /usr/bin/boinc, ldd gives a list of shared libraries the binary requires and the list will tell you which ones you have and which ones you don't have. It's possible Fedora updated and removed one of the shared libs boinc needs. In fact I would do the ldd command first before trying the other suggestions, I'm pretty sure a lib was deleted during an update. |
|
60)
Message boards :
BOINC Manager :
(debian) not connecting to client
(Message 41784)
Posted 25 Dec 2011 by (banished: ID 2516) Post: This a common problem but I need some info from you first before I can tell you how to fix it. Did you install BOINC from Debian repositories or did you download the Berkeley installer and install with that? |
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.