New Notfred diskless version out

Come join the... uh... er... fold.

Moderators: just brew it!, farmpuma

Re: New Notfred diskless version out

Postposted on Wed Aug 19, 2009 9:25 am

ethomaz wrote:I tested VirtualBox 3.0, but the current version doesn't support 2 or more core for 64bits Guests. :cry:


I'm using VirtualBox 3.0.4 r50677 on a windows Vista x64 machine with 8GB of mem. I set the memory for the VM at 1024MB, created a bridged virtual network adapter, and set the # of processors to 2 (this one is a E8500). It sees them just fine and my proc utilization on the host machine's task manager stays at 98-100%. I just pointed the VM to boot off of notfred's fold.iso I created on his website. (which is really cool that you can create a custom iso on the fly like that!)

It doesn't share processor time with the host machine as transparently as if you were just running the windows SMP client in the background but the computer is still usable. (unlike when I tried the same thing under VMware's free thing)
Image
Shinare
Gerbil XP
 
Posts: 352
Joined: Wed Jul 06, 2005 12:48 pm

Re: New Notfred diskless version out

Postposted on Wed Aug 19, 2009 3:14 pm

Shinare wrote:It doesn't share processor time with the host machine as transparently as if you were just running the windows SMP client in the background but the computer is still usable. (unlike when I tried the same thing under VMware's free thing)

What I do is to set the VM (either editing the .vmx file directly or through some GUI) to run at low priority. That way it should behave a bit better. Don't know how you do that in VirtualBox though.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24524
Joined: Mon May 24, 2004 2:19 am

Re: New Notfred diskless version out

Postposted on Wed Aug 19, 2009 4:36 pm

Ah, I'll have to look into that. I figured they did that by default and didn't even think about it.
Image
Shinare
Gerbil XP
 
Posts: 352
Joined: Wed Jul 06, 2005 12:48 pm

Re: New Notfred diskless version out

Postposted on Wed Aug 19, 2009 5:23 pm

Shinare wrote:Ah, I'll have to look into that. I figured they did that by default and didn't even think about it.

All Folding client "worker" processes, Linux or Windows or OS X, run in low priority. However, if you are running them inside of a VM, you have the VM processes to worry about as well, which is normal priority by default. Just more things to pay attention to.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24524
Joined: Mon May 24, 2004 2:19 am

Re: New Notfred diskless version out

Postposted on Thu Aug 20, 2009 12:31 pm

notfred wrote:Have you tried checking the box "Ignore asynchronous clocks" in Fahmon > Preferences > Monitoring tab? That should let it work with clients with different times on them.

Yes, I've had that set since my W2K-Barton decided to quit keeping time accurately (years ago). It doesn't work for this, for whatever reason.

On another topic, for those interested, it appears there's a way to run 4 cores in VMware.... on foldingforum.org they have a thread describing how you manually edit the .vmx config file to enable 4 cores per VM with 'numvcpus = "4"': Do Any VM's Use All 4 Cores?
Last edited by Ragnar Dan on Thu Aug 20, 2009 12:34 pm, edited 1 time in total.
Ragnar Dan
Gerbil Elder
Silver subscriber
 
 
Posts: 5355
Joined: Sun Jan 20, 2002 7:00 pm

Re: New Notfred diskless version out

Postposted on Thu Aug 20, 2009 12:32 pm

Shinare,

Sorry... I forgot to report that my OS is 32 bits (Windows XP).
Again, now correct: I tested VirtualBox 3.0, but the current version doesn't support 2 or more core for 64bits Guests in 32bits Hosts.

Ragnar Dan,

This only work with VMware Workstation installed and expires in 30 days without licence.
I'm brasilian and my English is bad, very bad.
ethomaz
Gerbil
 
Posts: 25
Joined: Wed Aug 20, 2008 9:00 am

Re: New Notfred diskless version out

Postposted on Thu Aug 20, 2009 1:53 pm

Ragnar Dan wrote:On another topic, for those interested, it appears there's a way to run 4 cores in VMware.... on foldingforum.org they have a thread describing how you manually edit the .vmx config file to enable 4 cores per VM with 'numvcpus = "4"': Do Any VM's Use All 4 Cores?



NICE! Just did that on my folding machine and now I only need 1 VM. I'm saving 1 GB of ram by getting rid of 1 of the 2 VMs I was running prior.
cygnus1
Gerbil
Gold subscriber
 
 
Posts: 65
Joined: Fri Oct 28, 2005 9:49 pm

Re: New Notfred diskless version out

Postposted on Thu Aug 20, 2009 2:13 pm

How can I change the timezone in my VM so the logged times are right? That would help FAHMON.
cygnus1
Gerbil
Gold subscriber
 
 
Posts: 65
Joined: Fri Oct 28, 2005 9:49 pm

Re: New Notfred diskless version out

Postposted on Thu Aug 20, 2009 3:19 pm

It may be possible to do something by changing option rtc.diffFromUTC in the virtual machine’s .vmx configuration file. I don't set it so it will get initialised at startup of the VM from my reading of the VMWare docs and it is in seconds. If you set it to 0 then it should run in UTC.
notfred
Grand Gerbil Poohbah
 
Posts: 3762
Joined: Tue Aug 10, 2004 10:10 am
Location: Ottawa, Canada

Re: New Notfred diskless version out

Postposted on Thu Aug 20, 2009 4:04 pm

I tried rtc.diffFromUTC = -14400. That sets the RTC to -4 hours (EST w/ DST). The timestamps in the log now shows current local time. We'll see when this WU is done and a new one is pulled down if the right download time is tracked in Fahmon and if it gives a proper estimate of PPD. I'll report back and let you know.
cygnus1
Gerbil
Gold subscriber
 
 
Posts: 65
Joined: Fri Oct 28, 2005 9:49 pm

Re: New Notfred diskless version out

Postposted on Thu Aug 20, 2009 9:22 pm

I think you actually want to set it to 0 as the FAHlog.txt timestamps are normally in UTC. Linux (which is what my diskless stuff uses) and every OS (apart from Windows) expects the RTC in UTC and not localtime.
notfred
Grand Gerbil Poohbah
 
Posts: 3762
Joined: Tue Aug 10, 2004 10:10 am
Location: Ottawa, Canada

Re: New Notfred diskless version out

Postposted on Mon Aug 24, 2009 2:13 pm

notfred wrote:I think you actually want to set it to 0 as the FAHlog.txt timestamps are normally in UTC. Linux (which is what my diskless stuff uses) and every OS (apart from Windows) expects the RTC in UTC and not localtime.


Yeah, windows sucks in that regard. It would be much simpler to do it the linux (and every other OS) way. To do everything internally with UTC and only fix it to local time when it's displayed. Basically what i was looking to do was to get the VM thinking my local time was UTC, so that my local time ended up in the logs and windows (and fahmon) would be play nice with the timestamps. It seemed to work, the log files showed my local time as the timestamps, but fahmon still showed a 4hr discrepancy.
cygnus1
Gerbil
Gold subscriber
 
 
Posts: 65
Joined: Fri Oct 28, 2005 9:49 pm

Re: New Notfred diskless version out

Postposted on Sat Aug 29, 2009 11:22 am

How hard would it be to add support for custom flags?
~~~~~ Caveman Make Fire!!! ~~~~~
caveman-jim
Gerbil
 
Posts: 12
Joined: Fri Dec 22, 2006 4:18 pm

Re: New Notfred diskless version out

Postposted on Sun Aug 30, 2009 4:34 pm

What do you mean by custom flags? Flags that you could specify at boot up time that get passed to the FAH client? I suppose I could add a CUSTOM="" option that would just get passed through to the client startup...
notfred
Grand Gerbil Poohbah
 
Posts: 3762
Joined: Tue Aug 10, 2004 10:10 am
Location: Ottawa, Canada

Re: New Notfred diskless version out

Postposted on Sun Aug 30, 2009 6:17 pm

Check your PM, please. 8)
~~~~~ Caveman Make Fire!!! ~~~~~
caveman-jim
Gerbil
 
Posts: 12
Joined: Fri Dec 22, 2006 4:18 pm

Re: New Notfred diskless version out

Postposted on Thu Sep 03, 2009 8:52 am

notfred,

The Diskless was geting many bad WUs, see link: http://foldingforum.org/viewtopic.php?f=19&t=11098. And I have to delete them to FAH get a new WU. How do I this without restart VM and edit vmdk???

If Diskless hasn't way to do this, then I will try to create a shell script to do this for me :wink:.
I'm brasilian and my English is bad, very bad.
ethomaz
Gerbil
 
Posts: 25
Joined: Wed Aug 20, 2008 9:00 am

Re: New Notfred diskless version out

Postposted on Fri Sep 04, 2009 8:42 am

It's been a request in the past, I just haven't got around to doing it yet.

You'll want to take a look at backup.sh and then go nuke all the places where it can save the backups, then reboot the system within the VM. There's already the reboot link, just need to go write the "nuke the backups" script.
notfred
Grand Gerbil Poohbah
 
Posts: 3762
Joined: Tue Aug 10, 2004 10:10 am
Location: Ottawa, Canada

Re: New Notfred diskless version out

Postposted on Fri Sep 04, 2009 10:40 am

notfred wrote:What do you mean by custom flags? Flags that you could specify at boot up time that get passed to the FAH client? I suppose I could add a CUSTOM="" option that would just get passed through to the client startup...


Specifically I was thinking the bigadv flag.
~~~~~ Caveman Make Fire!!! ~~~~~
caveman-jim
Gerbil
 
Posts: 12
Joined: Fri Dec 22, 2006 4:18 pm

Re: New Notfred diskless version out

Postposted on Tue Nov 03, 2009 7:24 pm

i am using VMware 6.5.3 with notfred, the clock in the VM is going to Slow, is there an setting i can put in that can fix this in the VMware setting file or notfred, as it makes estimating PPD with FAHmon not work correctly due to clock running to slow, if 2 VMs are running it goes even slower by 2x from what i can tell

on an QX6850 3.3Ghz, win7 64, it is set to high performance so speed step should be off (my AMD phenom II X4 system does not have this issue)
leexgx
Gerbil In Training
 
Posts: 2
Joined: Tue Nov 03, 2009 6:45 pm

Re: New Notfred diskless version out

Postposted on Tue Nov 03, 2009 10:08 pm

When I got home today I had an error (forgot what it was, but the VM was still folding). Upon restarting I have this error:

Code: Select all
SYSLINUX 3.51 0x466c807b EBIOS Copyright (C)1994-2007 H. Peter Anvin
Could not find kernel image: linux
boot:


I found a thread from a couple years ago with the same error, but I have been running fine since the weekend with the notfred image.

viewtopic.php?f=9&t=38164
notfred wrote:Hmm, it doesn't seem to be trying to retrieve the pxelinux.cfg file looking at the server logs, and the client shouldn't be trying to get anything called "linux", the file is called "kernel".

Make sure you actually have pxelinux.cfg folder which contains the file called default, and not a file called pxelinux.cfg. Also make sure all the options are exactly as shown in the diagrams, e.g. do you have the virtual root option on?



could someone help me with the steps to fix this?
frag85
Gerbil In Training
 
Posts: 1
Joined: Tue Nov 03, 2009 9:59 pm

Re: New Notfred diskless version out

Postposted on Sat Nov 07, 2009 9:45 pm

where do i edit to make -bigadv work ? (i see its the init file its in but does not save when its rebooted)
leexgx
Gerbil In Training
 
Posts: 2
Joined: Tue Nov 03, 2009 6:45 pm

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 10:42 am

Yesterday I noticed a new VM Appliance out
8 November 09: Updgard to kernel 2.6.30.1 and add VM time offset of 0.


I loaded it up on a dual core machine and got it folding with no problem. I don't see any change or difference in performance and wondered what the new version should by us?

I was also wondering if notfred had been following the issues with 2.10 and tear's "fix" to those performance problems. Would it be possible to incorporate that fix into the next version?

Thanks
weedacres
Gerbil
 
Posts: 18
Joined: Thu Nov 12, 2009 8:37 pm

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 12:34 pm

New kernel will support more hardware (especially more recent motherboard onboard ethernet controllers), it's not really a performance change. The VM time offset of 0 should help new installs in a VM work with the correct time zone to make monitoring in FAHMon (or similar) a lot easier - may not have to enable async clocks in FAHMon.

Can you point to the fix "tear" did and I'll see if I can include it in the next version. I just upgraded to Ubuntu 9.10 and it doesn't currently build (probably need to move up the glibc version) so the next version will fix the building issues.
notfred
Grand Gerbil Poohbah
 
Posts: 3762
Joined: Tue Aug 10, 2004 10:10 am
Location: Ottawa, Canada

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 1:59 pm

I think weed is referring to this: http://foldingforum.org/viewtopic.php?f=44&t=12011

It's been determined that adding MPICH_NO_LOCAL=1 environment variable makes the degradation go away for majority of users.


Have used it here and it is working.
GTX
Gerbil
 
Posts: 18
Joined: Sun Jun 11, 2006 9:39 pm

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 2:34 pm

Thanks! Looking at that thread it seems that it may be only useful for dual core. Should I enable it regardless or should I only do it if I find two CPUs?
notfred
Grand Gerbil Poohbah
 
Posts: 3762
Joined: Tue Aug 10, 2004 10:10 am
Location: Ottawa, Canada

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 2:41 pm

Very interesting. notfred: You have a quad core. Has your performance dropped significantly with the 2.10 core? I'm about to test it in my Linux VM on AMD dual core Opteron.
Ragnar Dan
Gerbil Elder
Silver subscriber
 
 
Posts: 5355
Joined: Sun Jan 20, 2002 7:00 pm

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 3:01 pm

To be honest I haven't been keeping a sharp lookout at it. I also run two copies of the SMP client. Given that I often see a core_a1 WU on it, I've given up chasing the max performance for a setup.
notfred
Grand Gerbil Poohbah
 
Posts: 3762
Joined: Tue Aug 10, 2004 10:10 am
Location: Ottawa, Canada

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 3:14 pm

notfred wrote:Thanks! Looking at that thread it seems that it may be only useful for dual core. Should I enable it regardless or should I only do it if I find two CPUs?


I would only enable when running two cpus. I don't think it has much if any affect folding on four cores.

When i tried it on my vm's running notfred's there was an increase in ppd but i didn't keep notes on how much.....
GTX
Gerbil
 
Posts: 18
Joined: Sun Jun 11, 2006 9:39 pm

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 3:21 pm

notfred wrote:To be honest I haven't been keeping a sharp lookout at it. I also run two copies of the SMP client. Given that I often see a core_a1 WU on it, I've given up chasing the max performance for a setup.

You had been rapidly catching up to me, but in the last number of weeks you are falling behind.

Anyway, my change doesn't seem to have made any difference. I stopped it in the middle of a WU and then restarted it with the environment variable change. I'll see how it does on the next WU since this one is nearly completed, but I'm doubtful.
Ragnar Dan
Gerbil Elder
Silver subscriber
 
 
Posts: 5355
Joined: Sun Jan 20, 2002 7:00 pm

Re: New Notfred diskless version out

Postposted on Fri Nov 13, 2009 5:05 pm

notfred wrote:New kernel will support more hardware (especially more recent motherboard onboard ethernet controllers), it's not really a performance change. The VM time offset of 0 should help new installs in a VM work with the correct time zone to make monitoring in FAHMon (or similar) a lot easier - may not have to enable async clocks in FAHMon.

Can you point to the fix "tear" did and I'll see if I can include it in the next version. I just upgraded to Ubuntu 9.10 and it doesn't currently build (probably need to move up the glibc version) so the next version will fix the building issues.

Thanks for the feedback.

GTX was correct about the forum link. http://foldingforum.org/viewtopic.php?f=44&t=12011
edit: There is also a very length discussion of the 2.10 problem here. http://foldingforum.org/viewtopic.php?f=44&t=11367

The 2.10 problem on 2 core machines is pretty much eliminated by tear's suggestion of either entering:
fold64 MPICH_LOCAL_NO=1 during the bootup or stopping folding activity and entering:
export MPICH_LOCAL_NO=1 and them resuming. I had some difficulty with that because I forgot to enter the -smp 4 in the command string as I was used to taking care of that in -configonly.

It works pretty much as advertised and eliminates most if not all of the 2.10 performance hit on 2 core folding. I also run it on my quads which are folding 4 cores. I always felt there was a small performance hit on the quads with 2.10 but I think it was more my imagination than anything else. It doesn't make any difference either way.
weedacres
Gerbil
 
Posts: 18
Joined: Thu Nov 12, 2009 8:37 pm

PreviousNext

Return to TR Distributed Computing Effort

Who is online

Users browsing this forum: Exabot [Bot] and 1 guest