removing all BOINC/SETI files ??

Questions and Answers : Unix/Linux : removing all BOINC/SETI files ??

Author	Message
Geoffrey Cowling Send message Joined: 24 Mar 00 Posts: 9 Credit: 3,391,118 RAC: 245	Message 946256 - Posted: 10 Nov 2009, 1:12:17 UTC
	When I run_manager the screen comes up, but EVERYTHING freezes (which it should not do on Linux!) ctrl-alt-delete ? no. All I can do is reset: sooo Windows. I suspect there are conflicting files: so I want to remove everything, and reinstall. (I have of course tried to do this, and apparently failed.) I have run seti for many years on Linux, and am running it on an old machine in the basement (upgraded to Ubuntu 9.10; no problems, even with Astropulse). I did a few things stupidly (with xkill!) and messed things royally. I could not reinstall with .sh, so ended up using apt-get (on Ubuntu 8.04). It started running, but only on 2 of my 3 cores--and no Astropulse. When a new version ( .sh) came out, I tried it, and ! it worked (but still only 2 cores, and no Astropulse.) I then upgraded to Ubuuntu 8.10. And now nothing but freeze. I am ignorant of where the boinc-client operates from (at boot? or only when manager runs?), and where the seti client resides-- but think if I can absolutely remove all files, I should be able to reinstall. I have removed the Ubuntu version, with apt and purge .. but --who knows? (I have looked at the difference in versions between boinc and BOINC!) any help lavishly appreciated Geoffrey ____________
	ID: 946256 ·

ML1 Volunteer tester Send message Joined: 25 Nov 01 Posts: 7212 Credit: 3,703,516 RAC: 681	Message 946575 - Posted: 11 Nov 2009, 20:06:45 UTC - in response to Message 946256. Last modified: 11 Nov 2009, 20:07:33 UTC
	When I run_manager the screen comes up, but EVERYTHING freezes (which it should not do on Linux!) ... When a new version ( .sh) came out, I tried it, and ! it worked (but still only 2 cores, and no Astropulse.) I then upgraded to Ubuuntu 8.10. And now nothing but freeze. All rather odd... Have you got a mix of having installed boinc from the Ubuntu package manager AND also installed locally a copy from the boinc downloads ...sh script? I would guess that multiple boincs running would get things a little confused... For the time being, you'll likely need to stay with the boinc 6.6.x series unless you're adventurous enough to manually upgrade the nVidia drivers to then be able to use the boinc 6.10.x series. Meanwhile, try uninstalling your present boinc, and then keep with the Ubuntu packaged version of boinc or look up the various instructions for installing boinc manually. The Ubuntu install installs boinc as a service. The ...sh script installs boinc into the directory BOINC wherever it is that you run the script. You must then start boinc manually also. For the cuda stuff to work, boinc must have read-write access to the /dev/nvidia* device files... And that's another story! (My trick is to set those device files group rw for the "video" group and then add the boinc user to the video group.) Hope that's helped a little... Good luck, Martin ____________ Mandriva Linux A user friendly OS! See new freedom Mageia2 The Future is what We make IT (GPLv3)
	ID: 946575 ·

Geoffrey Cowling Send message Joined: 24 Mar 00 Posts: 9 Credit: 3,391,118 RAC: 245	Message 947765 - Posted: 17 Nov 2009, 2:19:12 UTC - in response to Message 946575.
	Thank you for your reply--sorry mine has been delayed; a lot of work, a lot of crunching! I am sure you are right: there is a conflict mess somewhere. I think I have all Ubuntu removed (by dpkg) but not sure of the other. That is why I want to be sure I remove everything and start with clean slate. I have crunched since 2000 (on all kinds of machines) and had NO trouble, always using .sh script to install. My old computer in the basement is churning away (using 6.6.41 on Ubuntu 9.10), and even so underpowered is handling Astropulse. ( I am chary of using Ubuntu package, just because it only gave me 2 of my three cores, and no Astropulse. I have had no problems with the Bash script before). What I don't understand is where the actual seti@home program goes ...how it is started (I obviously know how to attach to the site), and whether there are any problems with dependencies for it .. CUDA is a 'nother ball game--I have tried and failed with it, but not anxious about it (and I have lost my eye candy, floppy windows and three D stuff in the mess--may get around to redoing that some time, but it is only seti that concerns me at the moment .. wasting cycles!) Thank you again -- a friendly note, as I tear out what is left of my hair, is always welcome :-) Geoffrey ____________
	ID: 947765 ·

Geoffrey Cowling Send message Joined: 24 Mar 00 Posts: 9 Credit: 3,391,118 RAC: 245	Message 948368 - Posted: 20 Nov 2009, 0:22:58 UTC - in response to Message 946575.
	OK -- I removed everything with boinc or BOINC in it -- and tried to reinstall 6.6.41. No joy. Same complete freeze on running manager. Removed everything again, tried 6.10. No problems, three cores (I now see this is set in preferences), and Astropulse. So while 6.6.41 runs on U 9.10. it wouldn't on my 8.10. Odd (I'll leave this up temporarily in case it helps anyone--and then I'll remove it) Linux still ought not completely freeze! Geoffrey ____________
	ID: 948368 ·

ML1 Volunteer tester Send message Joined: 25 Nov 01 Posts: 7212 Credit: 3,703,516 RAC: 681	Message 948502 - Posted: 20 Nov 2009, 13:42:36 UTC - in response to Message 948368. Last modified: 20 Nov 2009, 13:51:13 UTC
	Linux still ought not completely freeze! Quite so. The nVidia drivers are 'proprietary' and infiltrate the kernel despite being 'untrusted'. Hence why there is a requirement that those drivers announce themselves as such and the kernel declares itself "Tainted". You'll see that in the boot up logs. If something goes wrong in an untrusted driver in a tainted kernel, there's nothing much that can be easily debugged. The untrusted driver is proprietary. Who knows?! A strong design requirement for linux is that as much as possible is OUTSIDE of the kernel so that the kernel is as robust as possible. The latest versions of Xorg have moved yet more of the display code outside of the kernel for example. Meanwhile, Microsoft Windows has an unbelievable amount of application type code embedded into the Microsoft kernel including I believe all of the display code. One small error in any of that lot and you get the infamous BSOD. Or it just all freezes. Add into that the system level interception that is done by "antivirus" software that is to vet every system action... Quite a nightmare! All good fun! The extreme of using a small bug-free kernel has got to be the (highly parallel) GNU HURD system. However, development of that appears to be slow and the monolithic linux kernel is very successful and far less imperfect than certain other systems. Who knows... As we get to run ever more multiple processing cores, favour may shift to developing something similar to the GNU HURD, or even GNU HURD itself. Wikipedia GNU HURD GNU HURD home Happy crunchin', Martin (All just my own opinion as ever!) ____________ Mandriva Linux A user friendly OS! See new freedom Mageia2 The Future is what We make IT (GPLv3)
	ID: 948502 ·

Volunteer tester Send message Joined: 9 Apr 02 Posts: 12016 Credit: 19,625,122 RAC: 41,237	Message 948510 - Posted: 20 Nov 2009, 14:50:55 UTC - in response to Message 948502.
	Meanwhile, Microsoft Windows has an unbelievable amount of application type code embedded into the Microsoft kernel including I believe all of the display code. One small error in any of that lot and you get the infamous BSOD. Or it just all freezes. Windows Vista and all future Windows OSes have moved all display code to CPU Ring 3 thereby reducing the number of blue screens and preventing the entire system from crashing from a faulty display driver. This also allows the display driver on these OSes to be upgraded without a reboot. ____________
	ID: 948510 ·

Questions and Answers : Unix/Linux : removing all BOINC/SETI files ??