VMware Linux bigadv

Man of Honour
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
Ahh back to the good ole days.. before A3 on winders.

So the linux vmware image has been updated to 6.34 and A5 clients for Linux.

Ok so on my main windows (2600K) box I've managed to shave 2 minutes per frame off the average for a bigadv. Jumping it from 38K to 42K PPD and thats with my GTX460 running with zero effect on its output.

So everything you need is here: http://www.linuxforge.net/docs/crunching/fah-vmware.php

But a few tips.

1. Download the Zip not the 7z file as the 7z file is incomplete.
2. You will definitely need a process manager if you are running a Fermi card on the same box, if you are not then you don't.
3. Depending on your setup you may have problems with the Langouste client, I did so just set it up to download direct by turning off the proxy in the client.cfg after initial setup.
4. You *must* use the 3.0 version of Vmware Player otherwise you won't get the option to use 8 vcpus in the image. Vmware removed the option from later versions.


Affinity:
I'm using Bill2's Process Manager but you could use WinAFC or something, I just find Bill2's more intuitive to setup. :)
I also setup the Vmware image to use 7 threads instead of 8 so I save one for the GPU, your mileage may vary on this, and for my setup the jury is still out but for now its working, if I get some more time I will try different settings.

Any questions give me a holler in there. But free extra points aint bad.

And before anyone asks, yes I tried the VirtualBox setup on my dual Xeon box with 24 threads and it wasn't good. :( I don't think the VirtualBox setup is able to match the Windows client at this time
 
Last edited:
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
Note if you are using the latest version you don't need the fix below:

There is a fix for the Langouste error. It involves defining the USER variable before Langouste loads

At the console window:

root@FAH:~# _

Type: pico /etc/rc.d/rc.M

Then scroll down until you find the following lines:

Code:
# Start Langouste if enabled
. /etc/rc.d/rc.fah.conf
LPORT=8080
etc
etc

And before the first line add:

Code:
# Hack to define USER variable for Langouste
export USER=fah

So in the end it looks like this:

Code:
[B]# Hack to define USER variable for Langouste
export USER=fah
[/B]
# Start Langouste if enabled
. /etc/rc.d/rc.fah.conf
LPORT=8080
etc
etc
 
Last edited:
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
If the box is a dedicated folder not being used for anything else, then just use Linux, one of the boot CD's made from the same image as the VMWare images would probably be best.

Only reason to use VMware setup like this is if its a regular windows box that you want to optimise bigadv output on.
 
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
Yeah that will mess it up.

But sometimes you just get a normal smp WU, especially if you have had a few bigadv failures. I think you have to return at least 80% of all bigadv without failure or you will start to get normal smp's again.
 
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
Which "system" is locked to 112? You know the VM has a different mac address to the host machine right?

Have you tried opening a web browser and pointing it at the 118 address? You can shut down the folding from the Administration screen there.
 
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
I do yes, the folding system is locked to .112, always has been, the VMware address is .118. Pointing a browser to .118 it doesn't do anything, takes a while but nothing loads.

Simple things:

1. Can you ping .118 from the host?
2. Can you ping .112 from the VMWare console?
3. Can you ping or resolve anything from the VMWare console?
 
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
I keep having issues with HFM and the VM Image in that every now and again the status goes grey and I have to re-add the client in HFM (pointing at the same IP address as it was before!) to get it to show the correct status again. Not sure what is going on, but I can live with it.

Interesting, I'm not seeing anything like that. Both of you check the network adapter in the VMware image is set to Bridged mode.
 
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
Well that doesn't work for me unfortunately. It has well and truely screwed up, problem is I have zero idea when the unit is due to finish. So I can't delete everything and start again without the risk of losing all the work.

Just click on the console in the VMware and type:

cat unitinfo.txt

That will show you the client progress, or you can type:

cat FAHlog.txt
 
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
No thats normal

Check bridged mode in the VM Settings.

Then try turning off the windows firewall.
 
Last edited:
Man of Honour
OP
Joined
30 Oct 2002
Posts
15,807
Location
Surrey
FYI

linuxfah said:
I have uploaded a new VMware image to my server. I also have an updated native image built but have not had a chance to test that one out yet so that one I'll post up later. The latest image has the export USER set in the init script to allow for startup of Langouste. I was able to successfully submit a work unit using Langouste.

There are a number of other updates that can be found in my changelog here.

A new feature of this image is that the folding system now has the ability to output in JSON format, system usage/health, detailed folding stats, and TPF history for each client. The JSON interface should make it possible to make custom web pages, signatures, etc. off the folding system. Here is a small web report I put together that uses the JSON from my native folding rig and prints up some details and graphs.

http://www.linuxforge.net/fahstat
 
Back
Top Bottom