You are not logged in.
Hi all,
All three of my linux cpu clients are crashing now....
Launch directory: /home/simon/FAH
Executable: ./fah6
Arguments: -smp
[12:00:17] - Ask before connecting: No
[12:00:17] - User name: diesel1 (Team 45032)
[12:00:17] - User ID: 74700D444A448E6A
[12:00:17] - Machine ID: 1
[12:00:17]
[12:00:18] Loaded queue successfully.
[12:00:18] - Preparing to get new work unit...
[12:00:18] Cleaning up work directory
[12:00:18] + Attempting to get work packet
[12:00:18] Passkey found
[12:00:18] - Connecting to assignment server
fah6: ../sysdeps/unix/sysv/linux/getpagesize.c:32: __getpagesize: Assertion `_rtld_global_ro._dl_pagesize != 0' failed.
Segmentation fault
I am not sure what to do to solve this.
EDIT: Now the gpu client is also not working...
Initializing Nvidia gpu library
cudaMalloc CUDAStream::Allocate failed no CUDA-capable device is available
[12:47:25] mdrun_gpu returned
[12:47:25] Going to send back what have done -- stepsTotalG=0
[12:47:25] Work fraction=0.0000 steps=0.
[12:47:29] logfile size=4943 infoLength=4943 edr=0 trr=25
[12:47:29] + Opened results file
[12:47:29] - Writing 5481 bytes of core data to disk...
[12:47:29] Done: 4969 -> 1854 (compressed to 37.3 percent)
[12:47:29] ... Done.
[12:47:29] DeleteFrameFiles: successfully deleted file=work/wudata_04.ckp
[12:47:29]
[12:47:29] Folding@home Core Shutdown: UNSTABLE_MACHINE
It seems to have followed the xorg update.
Diesel1.
Last edited by diesel1 (2010-06-07 18:33:01)
Registered GNU/Linux user #140607.
Offline
diesel1,
I saw the same thing late last week. I'm guessing a new update broke something. When was the last time you ran pacman? Luckily I had a 64-bit Virtualbox image which I don't update much and it works fine.
Other relevant posts:
http://bbs.archlinux.org/viewtopic.php?id=98476
http://bbs.archlinux.org/viewtopic.php?id=60621
Right now I'm running:
GPU client in Arch (runs fine)
SMP client in ARCH VM (segfaults on real hardware after I reinstalled Arch)
SMP client in Windows 7 (runs fine)
The folding smp client depends on glibc, which had a recent update around the end of May.
Jim
Last edited by jimbok (2010-06-08 02:37:28)
Registered Linux User #402088
Offline
diesel1,
I saw the same thing late last week. I'm guessing a new update broke something. When was the last time you ran pacman?
-Jim
Eeeerrrrrr....maybe anytime, I have this crazy habit of running 'sudo pacman -Syu' a couple of times a day!
I just knew it was too good to be true - 2 gpu clients and 4 cpu clients running flat out and then KABOOM, all is dead - will I never learn to leave well alone.
So now I make my daughter's school lunch, walk the dog and maybe read the "other relevant posts".
TTFN,
Diesel1.
EDIT. NOOOOOoooooo, I just walked into the garage and automatically updated the last running gpu client! Maybe it is time I switched to Idiotbuntu!
Last edited by diesel1 (2010-06-07 22:54:23)
Registered GNU/Linux user #140607.
Offline
I was also getting segfaults. Apparently the new glibc changed the name of something which FAH shouldn't have been linking to directly (or something like that, I'm a bit fuzzy on the details).
Anyway, running
# /etc/rc.d/nscd start
before starting FAH fixed things for me. Apparently this provides a function which FAH uses instead of the glibc thing if available. It was also mentioned that nscd might not be completely reliable, so I'm not sure of all the repercussions in having nscd running.
Hopefully a new FAH binary is available soon so we don't need this workaround anymore.
Offline
I was also getting segfaults. Apparently the new glibc changed the name of something which FAH shouldn't have been linking to directly (or something like that, I'm a bit fuzzy on the details).
Anyway, running# /etc/rc.d/nscd start
before starting FAH fixed things for me. Apparently this provides a function which FAH uses instead of the glibc thing if available. It was also mentioned that nscd might not be completely reliable, so I'm not sure of all the repercussions in having nscd running.
Hopefully a new FAH binary is available soon so we don't need this workaround anymore.
Thanks for the tip, now my cpu and gpu clients are folding again.
Diesel1.
Last edited by diesel1 (2010-06-08 10:43:05)
Registered GNU/Linux user #140607.
Offline
Hi all,
Is it possible to run a hypervisor such as KVM/XEN (probably using virt-manager for gui) while still running cpu and gpu folding clients?
Diesel1.
Registered GNU/Linux user #140607.
Offline
Hi all,
Is it possible to run a hypervisor such as KVM/XEN (probably using virt-manager for gui) while still running cpu and gpu folding clients?
Diesel1.
Yes, I can only speak for Xen. You can run on Domain-0 and within the guests.
Offline
diesel1 wrote:Hi all,
Is it possible to run a hypervisor such as KVM/XEN (probably using virt-manager for gui) while still running cpu and gpu folding clients?
Diesel1.
Yes, I can only speak for Xen. You can run on Domain-0 and within the guests.
Cool.
Is it practical to convert an existing install to use as Dom-0 and from your experience what is the minimum practical total memory for a host system?
Quite a lot OT but please forgive me, I will stll be folding 99% of the time on the sytem.
Diesel1.
EDIT: Reading Xen wiki right now.
Last edited by diesel1 (2010-06-08 21:37:42)
Registered GNU/Linux user #140607.
Offline
The "fair share" CPU model in Xen is quite robust. You can read up on CreditScheduler.
For example.... if you only have dom0 and one VM running. No matter how much you set with "xm sched-credit" command, it will continue to utilize whatever many CPU cores is assigned to it @ 100%. So in this sense... if you convert your F@H instance from a VM up into dom0 would make no difference.
If you have multiple VMs running, then from the article clearly explains how CPU shares/priorities are sliced between the VMs. I do not know if dom0 will have higher priorities than all VMs in this case. General speaking, you would think so because dom0 have to handle all the actual/real system calls.
BTW, I just broke 10 million.
Offline
Congratulations on the 10 million!
Thanks for the info on xen.
Diesel1.
Registered GNU/Linux user #140607.
Offline
quad3d@work;
Congratulations on making the 10 Million milestone.
Cheers.
Pudge
Offline
Gratz, you have 755 times more points than me now!
Offline
I'm trying to get this going on my desktop but am having some trouble. I have the configuration set up according to the wiki but launching any of the daemons results in the following. Changed my user name, obviously.
bash: line 1: 9298 Segmentation fault nice --adjustment 19 /opt/fah/fah6 -verbosity 9 > /opt/fah/USER/myfah.log
I think this is something small and stupid because if I just run
/opt/fah/fah6 -verbosity 9
it seems to work fine. Is nice causing the segfault somehow?
edit: I should note, when I say just running the executable doesn't seg fault I mean literally just executing fah6. If I change the /etc/rc.d/foldingathome* files to remove nice, I still get the same segfault when launching the daemons.
edit2: So I have discovered the removing the output redirection allows things to run fine. I assume then, this must be some permissions problem? I have checked and all the /opt/fah*/USER directories are owned by the user as are all their contents including myfah.log so I'm still lost.
Last edited by crabmarkjellozest (2010-07-18 03:33:09)
Offline
There are many merits related to Arch Linux. The capability of Arch Linux distribution as well as people are big reasons to log on Arch Linux forums.
---
Moderator note: post edited due to possible spam. next attempt and you're gone.
Last edited by litemotiv (2010-08-15 14:25:40)
Offline
The Arch Linux folding team is now up to 14 millionaires.
Congratulations to neoxic for reaching the 1 million mile stone.
Cheers.
Pudge
Offline
Interesting. Does it require strong CPU power or strong GPU power?
Offline
Congratulations neoxic, well done on the 1 million points.
Diesel1.
Registered GNU/Linux user #140607.
Offline
Hi all,
After one of my servers bit the dustI thoought the stats might drop more than they have. So to reach 2 million points is more than I could have thought when I was pushing 300 PPd.
I have held off the 'sudo pacman -Syu' for a long time and lately therre seems to be many conflicts in the updrade, is anyone fully updated and running all clients?
All the best,
Diesel1.
EDIT: Typo.
Last edited by diesel1 (2010-08-21 00:44:14)
Registered GNU/Linux user #140607.
Offline
Interesting. Does it require strong CPU power or strong GPU power?
There are folding clients for both. The SMP client for multi-core CPUs, and a GPU client for the newer video cards. On most of my computers, I run both a SMP client to keep the CPU busy, and a GPU client.
GPU cards like the Nvidia 8800 GT, 8800 GTX, 9800 GT, and GTS250 have around 112 cores all of which are designed for number crunching. These are perfect for folding proteins and will fold a WU in about 3 hours.
Pudge
Offline
Hi all,
I have held off the 'sudo pacman -Syu' for a long time and lately therre seems to be many conflicts in the updrade, is anyone fully updated and running all clients?
All the best,
Diesel1.
EDIT: Typo.
You might want to hold back on a full system update. The current nvidia drivers, 256.44 has been hard locking some users screens, mine included. I reverted to 256.35 and that worked fine with the 2.6.34 kernel. I went ahead and installed the 2.6.35 kernel and I haven't been able to fold on my gpu since. Probably have to wait for an update nvidia driver from what I have been reading over at nvnews. I held the nvidia stuff back for a while, but I didn't want to get too far behind. I set up another cpu client to fill in until the nvidia stuff gets straightened out.
jimbok
Registered Linux User #402088
Offline
Hello everyone. This week i've decided to jump the arch linux folding band wagon. I figured it was a waste letting all those cores idle all day
Hi all,
After one of my servers bit the dustI thoought the stats might drop more than they have. So to reach 2 million points is more than I could have thought when I was pushing 300 PPd.
I have held off the 'sudo pacman -Syu' for a long time and lately therre seems to be many conflicts in the updrade, is anyone fully updated and running all clients?
All the best,
Diesel1.
EDIT: Typo.
I've run both the normal and the smp client on a fully updated system. Everything works. I'll try out the nvidia gpu client later but as jimbok already noticed, there have been many complaints about the latest nvidia driver. I've been issue free however, so it will probably work for me.
Offline
Interesting. Does it require strong CPU power or strong GPU power?
Strong power isn't needed, I'm running two computers with 1.66+668 Ghz, they're idling all day anyway so why not put them to good use?
Although I bought a "new" gfx card (radeon hd 3650) recently, I might start a GPU client on it.
Offline
Strong power isn't needed
Indeed, I'm on an Intel Atom 1.6 GHz here. It isn't fast, but at least it does something between two keystrokes.
I'm running two computers with 1.66+668 Ghz
Gulp! 668 GHz? Do you work for CERN or NASA? I suppose you meant 668 MHz.
Offline
Didn't read through the whole thread but I want to support F@H.
I only have a Sempron 2Ghz. The AUR Package says F@H is for i686 only. Any chance to have it running on a x86_64 installation (the SMP won't go well with my CPU, will it?)
Last edited by Barghest (2010-09-05 10:51:21)
Offline