Help me build a Folding Farm

CharlieChan · Post by **CharlieChan** » Sun Dec 07, 2003 8:57 am

David:

Well Done!

I could advise you on what to do next but first I need to know what type of installation you used to install redhat 9.0. Was it desktop, workstation, server, or custom? Did you choose the default installation or added further packages? Also have you updated you redhat installation by using up2date - this is probably not a good option as it will probably break the drivers you installed to get the nforce board to work.

Does your existing internal network use the address 192.168.0. ? The default installation of ltsp use this address and it would save you time changing all the config files on ltsp if you internal do not use this address range.

Charlie.

haysdb · Post by **haysdb** » Sun Dec 07, 2003 2:39 pm

I won two battles. I have the feeling the war will be long.

David

haysdb · Post by **haysdb** » Sun Dec 07, 2003 3:08 pm

CharlieChan wrote:David:

Well Done!

I could advise you on what to do next but first I need to know what type of installation you used to install redhat 9.0. Was it desktop, workstation, server, or custom? Did you choose the default installation or added further packages? Also have you updated you redhat installation by using up2date - this is probably not a good option as it will probably break the drivers you installed to get the nforce board to work.

Does your existing internal network use the address 192.168.0. ? The default installation of ltsp use this address and it would save you time changing all the config files on ltsp if you internal do not use this address range.

Charlie.

Charlie,

I did an "Other" installation and installed basically anything I thought I even remotely might use. I figured that was safer than trying to figure out how to install things after the fact. A lot of "crap" got installed, but I'm assuming I can just "turn off" what I don't need.

I actually did try to connect to RHN, but it sort of "went out to lunch" on the license agreement page. I actually have a subscription to RHN which is probably just about to expire, so if there are some things I need to do, I should do it ASAP.

I created the nvnet driver from the sources, so I think I could manage to upgrade a new kernel if I had to.

My existing network does use the 192.168.0. range. It would be easy enough to change, except for the Dlink router, which I think wants to be 192.168.0.1. [Edit]No, not true, the LAN page in my router configuration allows me to specify the IP address of the router.[/Edit]

Should I install a second NIC in the FAH Server and put the FAH clients on a subnet? My router does have the ability to assign static IP addresses by MAC Address, so I think I can get by without doing this, but I will if it makes sense to do so.

David

CharlieChan · Post by **CharlieChan** » Sun Dec 07, 2003 4:58 pm

David,

The farm has to be on its own network becauce the dhcp server has to point the client to the location of the kernel on the farm server. I would not update the server as upgrades usually patch security issues which is not a problem as you are already behind a firewall. My advise would be to change you current network to 192.168.1.0, get a second network card for the farm server. Assign one of the network card to 192.168.1. and the other to 192.168.0. Check you have the following packages installed using rpm -q [package name]
tftp-server
dhcp
nfs-utils
portmap
dnsmasq - get this here.

Remove bind using rpm -e as dnsmasq is a simple bind replacement which does not need to be config. You can then download the following from ltsp,
ltsp_core...rpm
ltsp_kernel...rpm
ltso_local_app...rpm
pxestuff_3.0.5...tgz

At this point you are going to need a client to contiune.
Remember to use fixed IP addresses for the farm server.

Charlie.

haysdb · Post by **haysdb** » Mon Dec 08, 2003 9:07 am

Does this sound about right?

Network configuration:

Change the IP address of my Dlink broadband router (the DHCP server on my current network) from 192.168.0.1 to 192.168.1.1, i.e. move it from subnet 0 to 1.
Update the router DHCP configuration to assign 192.168.1.x numbers to each of my Windows machines.
Update each of my Windows clients to point to the new 102.168.1.1 address for their DHCP server.
Point the built-in NIC in the farm server to 192.168.1.1 to make it part of the existing network.
Install a second NIC in the Farm Server to handle the network of farm servers, which will be assigned addresses in the 192.168.0.x range.

David

haysdb · Post by **haysdb** » Mon Dec 08, 2003 9:24 am

Looks like I will order up the bits for at least one client this evening, along with a network card for the server.

Charlie, you mentioned Intel Pro network cards, but those were for the clients, right? Can the card for the server be any old card, or are there network cards that work better with Linux?

David

CharlieChan · Post by **CharlieChan** » Mon Dec 08, 2003 10:49 am

David,

You are moving in the right direction. I expect you will have at least one client working before the end of the week.

haysdb wrote: Charlie, you mentioned Intel Pro network cards, but those were for the clients, right? Can the card for the server be any old card, or are there network cards that work better with Linux?

Most will work with Linux, if you stick to well known brands you will probably find it is supported already. Ones I used for my farm server are netgear FA311. The reason I used Intel is they are well supported on Linux and cheap on ebay, I brought a box of 10 for £50 - £5 each or about $9.

Charlie.

haysdb · Post by **haysdb** » Mon Dec 08, 2003 8:25 pm

OK, the bits for the first folding client are:

Mobo: Biostar M7VIZ ($59)
CPU: 2500+ Barton ($86)
HSF: AlphaPAL (on hand)
RAM: 256MB Apacer (no-name) PC-2700 ($38)
PSU: 300W Zalman PSU (on hand)

And the pieces needed to establish the network:

Hub: Zonet 8-port 10/100 switch ($25)
NIC: Dlink DFE-530TX+ ($10) for farm server

David

haysdb · Post by **haysdb** » Mon Dec 08, 2003 8:41 pm

I'm going to see about switching my existing network over to the .1 subnet tonight. That should be trivial.

Then I'll see about installing the software Charlie listed above.

I'm seriously considering going ahead and bringing RedHat 9 current with all the patches. I agree that it's probably not necessary, but I'd just feel better about it, even if it causes me some setbacks in terms of things breaking.

David

haysdb · Post by **haysdb** » Thu Dec 11, 2003 2:48 am

CharlieChan wrote:Check you have the following packages installed using rpm -q [package name]
tftp-server
dhcp
nfs-utils
portmap
dnsmasq - get this here.

OK, these packages are all installed:

Code: Select all

# rpm -q tftp-server
tftp-server-0.32-4
# rpm -q dhcp
dhcp-3.0pl1-23
# rpm -q nfs-utils
nfs-utils-1.0.1-3.9
# rpm -q portmap
portmap-4.0-54
# rpm -q dnsmasq
dnsmasq-1.18-2

Remove bind using rpm -e as dnsmasq is a simple bind replacement which does not need to be config.

bind has been uninstalled:

Code: Select all

# rpm -q bind
package bind is not installed

You can then download the following from ltsp,
ltsp_core...rpm
ltsp_kernel...rpm
ltsp_local_app...rpm
pxestuff_3.0.5...tgz

The three rpm's are downloaded and installed:

Code: Select all

# rpm -q ltsp_core
ltsp_core-3.0.9-0
# rpm -q ltsp_kernel
ltsp_kernel-3.0.12-0
]# rpm -q ltsp_local_apps
ltsp_local_apps-3.0.0-0

I unpacked the pxestuff file and per the README file:

Moved the bzImage and initrd files to the /tftpboot/lts directory
Moved the pxelinux.0 bootloader to the /tftpboot/lts directory
Moved the pxelinux.cfg/default file to /tftpboot/lts/pxelinux.cfg/default

Does this look right?

Code: Select all

# ll /tftpboot/lts/*
-rw-r--r--    1 root     root       679237 Sep  5  2002 bzImage-2.4.19-ltsp-1
-rw-r--r--    1 root     root      2048000 Sep  5  2002 initrd-2.4.19-ltsp-1
-rw-r--r--    1 fah      fah          9952 Apr  5  2002 pxelinux.0
-r--r--r--    1 root     root      1718784 Nov  1 22:16 vmlinuz-2.4.22-ltsp-2
 
/tftpboot/lts/2.4.22-ltsp-2:
total 1704
-rw-r--r--    1 root     root       750592 Dec 11 03:37 bzImage-2.4.22-ltsp-2
-rw-r--r--    1 root     root       962560 Dec 11 03:37 initrd-2.4.22-ltsp-2.gz
-rw-r--r--    1 root     root         9952 Nov  1 22:16 pxelinux.0
drwxr-xr-x    2 root     root         4096 Dec 11 03:36 pxelinux.cfg
-rw-r--r--    1 root     root         3287 Nov  1 22:16 README.pxe
 
/tftpboot/lts/pxelinux.cfg:
total 4
-rw-r--r--    1 fah      fah           131 Sep  5  2002 default

It's here that things get fuzzy.

The pxe README file talks about adding lines to the /etc/dhcpd.cong file, but I don't have one. It also says to restart dhcp. I think maybe it's not running, or is this what dnsmasq replaced? Like I said, fuzzy.

At this point you are going to need a client to continue.
Remember to use fixed IP addresses for the farm server.

Can I just let the NIC attached to the main network pick up a dynamic IP address from the DHCP server in the broadband router? If it must be a fixed IP address, why must it be fixed? Is it because the two NICs must know each other's IP address?

david

CharlieChan · Post by **CharlieChan** » Thu Dec 11, 2003 4:15 am

David,

Now the fun starts

.

First initialize the ltsp installation.

cd /opt/ltsp/templates
./ltsp_initialize

Next create a dhcpd.conf in /etc, here is mine

Code: Select all

# Global Parameters

option routers			192.168.2.200;
option subnet-mask		255.255.255.0;
option domain-name		"sandpit.100akerwood.net";
option domain-name-servers	192.168.2.200;
option broadcast-address	192.168.2.255;
option root-path		"192.168.2.200:/opt/ltsp/i386";
option option-128 code 128 = string;
option option-129 code 129 = text;

default-lease-time 86400;

ddns-update-style ad-hoc;

subnet 192.168.2.0 netmask 255.255.255.0 {
	range 192.168.2.101 192.168.2.151;

	group {
		use-host-decl-names	on;
		option log-servers	192.168.2.200;

	        host ws001 {
        	        hardware ethernet	00:90:27:8E:05:67;
                	fixed-address 		192.168.2.101;
#			filename		"/lts/vmlinuz-2.4.21-ltsp-1";
			filename		"/lts/pxelinux.0";
			option vendor-encapsulated-options 09:0f:80:00:0c:4e:65:74:77:6f:72:6b:20:62:6f:6f:74:0a:07:00:50
:72:6f:6d:70:74:06:01:02:08:03:80:00:00:47:04:80:00:00:00:ff;
		}

	        host ws002 {
        	        hardware ethernet	00:90:27:BE:B6:4E;
                	fixed-address 		192.168.2.102;
			#filename		"/lts/vmlinuz-2.4.21-ltsp-1";
			filename		"/lts/pxelinux.0";
			option vendor-encapsulated-options 09:0f:80:00:0c:4e:65:74:77:6f:72:6b:20:62:6f:6f:74:0a:07:00:50:
72:6f:6d:70:74:06:01:02:08:03:80:00:00:47:04:80:00:00:00:ff;
		}

	        host ws003 {
        	        hardware ethernet	00:90:27:BE:BA:B0;
			fixed-address 		192.168.2.103;
			#filename		"/lts/vmlinuz-2.4.21-ltsp-1";
			filename		"/lts/pxelinux.0";
			option vendor-encapsulated-options 09:0f:80:00:0c:4e:65:74:77:6f:72:6b:20:62:6f:6f:74:0a:07:00:50:
72:6f:6d:70:74:06:01:02:08:03:80:00:00:47:04:80:00:00:00:ff;
		}

	        host ws004 {
        	        hardware ethernet	00:90:27:8E:04:D2;
			fixed-address 		192.168.2.104;
			#filename		"/lts/vmlinuz-2.4.21-ltsp-1";
			filename		"/lts/pxelinux.0";
			option vendor-encapsulated-options 09:0f:80:00:0c:4e:65:74:77:6f:72:6b:20:62:6f:6f:74:0a:07:00:50
:72:6f:6d:70:74:06:01:02:08:03:80:00:00:47:04:80:00:00:00:ff;
		}

	        host ws005 {
        	        hardware ethernet	00:C0:A8:F6:00:23;  #00:90:27:BE:B5:D1;
			fixed-address 		192.168.2.105;
			#filename		"/lts/vmlinuz-2.4.21-ltsp-1";
			filename		"/lts/pxelinux.0";
			option vendor-encapsulated-options 09:0f:80:00:0c:4e:65:74:77:6f:72:6b:20:62:6f:6f:74:0a:07:00:50:
72:6f:6d:70:74:06:01:02:08:03:80:00:00:47:04:80:00:00:00:ff;
		}

	        host ws006 {
        	        hardware ethernet	00:02:B3:36:51:F6;
			fixed-address 		192.168.2.106;
			#filename		"/lts/vmlinuz-2.4.21-ltsp-1";
			filename		"/lts/pxelinux.0";
			option vendor-encapsulated-options 09:0f:80:00:0c:4e:65:74:77:6f:72:6b:20:62:6f:6f:74:0a:07:00:50:
72:6f:6d:70:74:06:01:02:08:03:80:00:00:47:04:80:00:00:00:ff;
		}

	        host ws007 {
        	        hardware ethernet	00:02:B3:36:52:17;
			fixed-address 		192.168.2.107;
			#filename		"/lts/vmlinuz-2.4.21-ltsp-1";
			filename		"/lts/pxelinux.0";
			option vendor-encapsulated-options 09:0f:80:00:0c:4e:65:74:77:6f:72:6b:20:62:6f:6f:74:0a:07:00:50:
72:6f:6d:70:74:06:01:02:08:03:80:00:00:47:04:80:00:00:00:ff;
		}


	}

}

You can use this file, just replace IP of you farm server and subnet. Notice mine is on 192.168.2.0.

Here is hosts file in /etc

Code: Select all

127.0.0.1	localhost.localdomain	localhost

192.168.2.200	hunnytree.sandpit.100akerwood.net	hunnytree

192.168.2.101	ws001.sandpit.100akerwood.net	ws001
192.168.2.102	ws002.sandpit.100akerwood.net	ws002
192.168.2.103	ws003.sandpit.100akerwood.net	ws003
192.168.2.104	ws004.sandpit.100akerwood.net	ws004
192.168.2.105	ws005.sandpit.100akerwood.net	ws005
192.168.2.106	ws006.sandpit.100akerwood.net	ws006
192.168.2.107	ws007.sandpit.100akerwood.net	ws007

Next, /etc/exports

Code: Select all

#
## LTS-begin ##

#
# The lines between the 'LTS-begin' and the 'LTS-end' were added
# on: Fri Jul 11 15:49:50 BST 2003 by the ltsp installation script.
# For more information, visit the ltsp homepage
# at http://www.ltsp.org
#

/opt/ltsp/i386                  192.168.2.200/255.255.255.0(ro,no_root_squash,sync)
/var/opt/ltsp/swapfiles         192.168.2.200/255.255.255.0(rw,no_root_squash,async)

#
# The following entries need to be uncommented if you want
# Local App support in ltsp
#
/home/fah/ws001	192.168.2.200/255.255.255.0(rw,no_root_squash,sync)
/home/fah/ws002	192.168.2.200/255.255.255.0(rw,no_root_squash,sync)
/home/fah/ws003	192.168.2.200/255.255.255.0(rw,no_root_squash,sync)
/home/fah/ws004	192.168.2.200/255.255.255.0(rw,no_root_squash,sync)
/home/fah/ws005	192.168.2.200/255.255.255.0(rw,no_root_squash,sync)
/home/fah/ws006	192.168.2.200/255.255.255.0(rw,no_root_squash,sync)
/home/fah/ws007	192.168.2.200/255.255.255.0(rw,no_root_squash,sync)

## LTS-end ##

Next, /opt/ltsp/i386/etc/lts.conf

Code: Select all

#
# Config file for the Linux Terminal Server Project (www.ltsp.org)
#

[Default]
        SERVER             = 192.168.2.200
        XSERVER            = auto
	X_MOUSE_PROTOCOL   = "PS/2"
	X_MOUSE_DEVICE     = "/dev/psaux"
	X_MOUSE_RESOLUTION = 400
	X_MOUSE_BUTTONS    = 3
	USE_XFS            = N
	LOCAL_APPS         = Y
        RUNLEVEL           = 3
	RCFILE_01          = startfah1


#------------------------------------------------------------------------------
#
# Example of specifying X settings for a workstation
#
[ws006]
        XSERVER            = auto
        LOCAL_APPS         = Y
        USE_NFS_SWAP       = N
        SWAPFILE_SIZE      = 64m
        RUNLEVEL           = 3
	RCFILE_02          = startfah2

[ws007]
        XSERVER            = auto
        LOCAL_APPS         = Y
        USE_NFS_SWAP       = N
        SWAPFILE_SIZE      = 64m
        RUNLEVEL           = 3
	RCFILE_02          = startfah2

#
# ws004 is my virtual workstation running in a VMware session
#
#[ws004]
#        DNS_SERVER         = 192.168.0.254
#        XSERVER            = auto
#        X4_BUSID           = "PCI:0:15:0"
#        X_MODE_0           = 800x600
#        LOCAL_APPS         = N
#        USE_NFS_SWAP       = N
#        SWAPFILE_SIZE      = 64m
#        RUNLEVEL           = 5

#------------------------------------------------------------------------------
#
# Example of a workstation configured to load some modules
#
#[ws001]
#    MODULE_01       = agpgart.o		# This is for i810 video
#    MODULE_02       = uart401.o
#    MODULE_03       = sb.o io=0x220 irq=5 dma=1
#    MODULE_04       = opl3.o

#------------------------------------------------------------------------------
#
# Example of ws001 configured for local apps
#
#[ws001]
#    LOCAL_APPS      = Y
#    LOCAL_WM        = Y
#    NIS_DOMAIN      = ltsp
#    NIS_SERVER      = 192.168.0.254

#------------------------------------------------------------------------------
#
# Example of a serial printer attached to /dev/ttyS1 on workstation ws001
#
#[ws001]
#    PRINTER_0_DEVICE   = /dev/ttyS1
#    PRINTER_0_TYPE     = S		# P-Parallel, S-Serial
#    PRINTER_0_PORT     = 9100		# tcp/ip port: defaults to 9100
#    PRINTER_0_SPEED    = 9600		# baud rate: defaults to 9600
#    PRINTER_0_FLOWCTRL = S		# Flow control: S-Software (XON/XOFF),
#					# H-Hardware (CTS/RTS)
#    PRINTER_0_PARITY   = N		# Parity: N-None, E-Even, O-Odd
#					# (defaults to 'N')
#    PRINTER_0_DATABITS = 8		# Databits: 5,6,7,8 (defaults to 8)

Next, /opt/ltsp/i386/etc/rc.local

Only a couple of changes in this large script

Code: Select all


DEFAULT_SERVER=`get_cfg SERVER 192.168.2.200'

################################################################################
#
# Mount filesystems
#

if [ "${LOCAL_APPS}" = "Y" ]; then
    pr_set 85 "Mounting additional filesystems"
    echo "Mounting additional filesystems..."
    mount -t nfs -o nolock ${NFS_SERVER}:/home/fah/${HOSTNAME} /home/fah
fi

I have not list the file here as it is big. I only shown the changes needed.

Next, /opt/ltsp/i386/etc/rc.d/startfah1

Code: Select all

#!/bin/bash

#
# This is a sample rc script for LTSP.
#
# To run this script, add a RCFILE_xx entry to the lts.conf file,
# where 'xx' is a 2-digit number between 01 and 10.
# For example, if this script were called 'start_nbd', you
# would add an entry like this:
#
#    RCFILE_01 = start_nbd
#
echo "Starting F@H"
cd /home/fah/CPU1/
nohup ./FAH3.25-Linux-beta.exe -forceasm -advmethods > /dev/null&

Your do not needed the startfah2 file as it starts the second F@H client for dual CPU.

Copy the exe nohup, nice from redhat 9 to /opt/ltsp/i386/bin

You will need to create the following directories

mkdir /opt/ltsp/i386/home/fah
mkdir /home/fah/ws001
....
mkdir /home/fah/wsxxx

In each of the wsXXX directories put a client.cfg and folding client. I would advise using 3.25 beta as ltsp does not use the new glibc.

Now boot a client and see what happens

.

Charlie.

haysdb · Post by **haysdb** » Thu Dec 11, 2003 5:09 pm

I won't get to the configuration until later this evening. The 'client' was shipped yesterday, so it may be next week before I get anything running.

Setting this up is definitely not for the weak. This is not something that just any moron can do. With Windows, all the magic is hidden. With Linux, you see all the magic, but don't understand any of it at first, so it's overwhelming. You read things like "restart DCHP" without the foggiest notion of how one would go about doing that. One is bombarded with acronyms: DHCP, TFTP, LTSP, NFS, DNS, NAT, PXE, glibc, rpm. The editors with which the configuration files are updated are non-intuitive. The learning curve is either freightening or wonderful, depending on how one looks at it.

"Stable" and "support" are words not used lightly in the same sentence with "Linux". Support comes from a hundred different places, a whole pocketful for each "package" that one comes in contact with. However, it's never really a lack of information or help that's the problem, but knowing WHERE to look, WHO to ask, and being able to absorb, sort, and collate millions of bits of information cleaned from a thousand different sources.

I'm still trying to decide is this is a GIGANTIC P-I-T-A, or whether it's THE MOST FUN I'VE HAD SINCE COLLEGE.

I have more or less decided that the Linux server won't necessarily be stuck in the closet and forgotten about when I'm finished setting up the farm. It's nice having it physically in my office where I can toggle between it and my desktop Windows machine. The problem with that is it's a noisy Shuttle XPC. And I had JUST completed an upgrade to my office machine to make it QUIET. It may not be long before I build a quiet Linux machine for the office that I can just PLAY WITH (install Fedora), and move the XPC Farm Server to the basement.

I'm just rambling.

David

haysdb · Post by **haysdb** » Thu Dec 11, 2003 5:13 pm

Charlie,

Could you edit your post containing the long lines in the CODE block beginning
09:0f:80:00:0c:4e:65:74:77:6f:72:6b
and manually wrap the lines? I think that's what's causing all the other posts on this page not to wrap to my window width, forcing me to scroll left and right to read them.

Thanks!

BTW, any idea what this string of hex digits is? The "encapsulation options" line?

David

ColdFlame · Post by **ColdFlame** » Thu Dec 11, 2003 5:36 pm

<FLAMESUIT ON>To answer your Linux rant David - yes I think Linux is complex if you want to do something non-trivial and this is why I laugh when people say Linux will kill Windows soon.</FLAMESUIT OFF>

Now, I realize that Linux is the best solution for folding PCs because you don't have to pay for Windows license but this is the only advantage to me at this point. I might be blind though

CharlieChan · Post by **CharlieChan** » Thu Dec 11, 2003 6:12 pm

haysdb wrote: BTW, any idea what this string of hex digits is? The "encapsulation options" line?

I guesting here, I think it is the boot loader. When a PC boots the BIOS tells the hardware were the boot loader is on the hard drive which then loads the OS. When the PC network boots the network card search for a dhcp server using the MAC address. The server sends the boot loader which then point to the OS. Most of the stuff I posted here are from this guy. The instructions are easy to follow if you know a little about Linux but I imagine it can be a little overwhelming for someone used to Windows. However, Windows was not always this easy to used. I still remembers the days of Windows 3.0 and MS DOS 5. In those days you have to tweak file after file to get a program or a driver to work, oops showing my age

Charlie.

ColdFlame · Post by **ColdFlame** » Thu Dec 11, 2003 7:27 pm

I still remember DOS 3.30 the best OS ever

haysdb · Post by **haysdb** » Fri Dec 12, 2003 5:08 am

Charlie,

Tell me again why I should put the folding clients on the 198.168.0 subnet?

The reason I am questioning this is that it would make more sense to me to put it on, say, the 192.168.2 subnet. It's just a stupid mental block with me. What is it I will have to configure if I don't use the 0 subnet? Was it dnsmasq?

And speaking of dnsmasq. What does this replace? It replaces bind, but nowhere in any of the documents I have read on setting up folding farms does anyone talk about the need to configure bind, or dns.

Also speaking of dnsmasq, I am getting some strange errors. One was a command is_yes in /etc/rc5.d/S14dnsmasq. I did a Google search for is_yes and came up empty as far as it being a shell built-in or Linux command. I'm getting similar errors on startup and shutdown. The messages flash by quick and I haven't gotten them written down. Do those startup messages get written to a log?

Here. I'm not positive, but I think the errors are in one of these lines. What are these? Commands that are missing from my system? Some part of dnsmasq that didn't get installed?

msg_starting
msg_network_down
msg_already_running
msg_not_running
msg_stopping
msg_usage

[Edit]
I think I can partially answer my own question. These, along with is_yes, are functions expected to be defined in /etc/rc.d/init.d/functions. RedHat doesn't define these functions, so I either need to find a source for them, replace them in the dnsmasq scripts, or write functions to do what I think they do. I'd much rather just find the source for these functions. I'm sure they are only a few lines each.
[/Edit]

[Edit]
I originally installed the binary. To make sure I had the latest, I downloaded the dnsmasq "tarball", rebuilt the .rpm, and did an rpm -Uhv --force to install the "older" version. Somehow the binary was 1.18-2 and the last source is 1.18-1. The script in the 'source' version does not use these functions.
[/Edit]

David

CharlieChan · Post by **CharlieChan** » Fri Dec 12, 2003 7:32 am

haysdb wrote: Tell me again why I should put the folding clients on the 198.168.0 subnet?

If you have run the ltsp initialization script, you will notice it has changed about 6 files. Most of these files defaults to 192.168.0. network, if you wish to move subnet you will have to change the setting in these files. My advise would be to do so after you have got the farm working.

And speaking of dnsmasq. What does this replace? It replaces bind, but nowhere in any of the documents I have read on setting up folding farms does anyone talk about the need to configure bind, or dns.

If you follow my earlier link you will see on the second page the author refers to bind. The setup you are employing for the farm is not a proper subnet, in fact you farm is behind a NAT. You farm server is the router and NAT not a bridge, thats why you need dns on the server. The setup is not very elegant but it works.

I originally installed the binary. To make sure I had the latest, I downloaded the dnsmasq "tarball", rebuilt the .rpm, and did an rpm -Uhv --force to install the "older" version. Somehow the binary was 1.18-2 and the last source is 1.18-1. The script in the 'source' version does not use these functions.

I take it dnsmasq now works.

Charlie.

Lockheed · Post by **Lockheed** » Fri Dec 12, 2003 9:11 am

Post removed

haysdb · Post by **haysdb** » Fri Dec 12, 2003 9:23 am

CharlieChan wrote:If you have run the ltsp initialization script, you will notice it has changed about 6 files. Most of these files defaults to 192.168.0. network, if you wish to move subnet you will have to change the setting in these files. My advise would be to do so after you have got the farm working.

I'm waiting on the parts for the first client machine so I have a little free time.

I have identified the following files.

/etc/dhcpd.conf
/etc/hosts
/etc/exports
/etc/hosts.allow
/etc/samba/smb.conf
/opt/ltsp/i386/etc/rc.local
/opt/ltsp/i386/etc/lts.conf

david

Edit: Added the two files Lockheed mentions in the next post, so that I have the full list all in one place.

Lockheed · Post by **Lockheed** » Fri Dec 12, 2003 9:29 am

Post removed

haysdb · Post by **haysdb** » Fri Dec 12, 2003 9:57 am

OK, there should only be one DHCP server on a network, and it has to be in the farm server, so I would have to turn DHCP off in the broadband router. But that would then require me to manually configure each of my Windows machines, not to mention a ReplayTV which picks up an IP address via DHCP. That might be easier in the long run, but right now that doesn't sound like a desireable solution.

I will need a second switch anyway, and the extra NIC I bought cost only $10, so other than a bit of extra configuration, what's the downside of a separate network?

David

Lockheed · Post by **Lockheed** » Fri Dec 12, 2003 10:20 am

Post removed

CharlieChan · Post by **CharlieChan** » Sat Dec 13, 2003 2:11 pm

haysdb wrote:OK, there should only be one DHCP server on a network, and it has to be in the farm server, so I would have to turn DHCP off in the broadband router. But that would then require me to manually configure each of my Windows machines, not to mention a ReplayTV which picks up an IP address via DHCP. That might be easier in the long run, but right now that doesn't sound like a desireable solution.

The dhcpd.conf I listed will allocated IP in the range 192.168.2.101 to 192.168.2.151 for any machine on the network. If you like to used the network range 192.168.0. just alter all the IP's in the files you listed. You did say you wanted the farm to be in the basement in which case you would probably need to put it on a subnet.

Charlie.

Lockheed · Post by **Lockheed** » Sun Dec 14, 2003 1:38 pm

Post removed

CharlieChan · Post by **CharlieChan** » Sun Dec 14, 2003 2:52 pm

Lockheed wrote: Why does it matter where the machine is located? I have machines all over the house, all on the same subnet.

I am not a network expert, this is just a hobby, my understanding is for a group of computer to exist on the same subnet they must all be connected to a common switch. If the building housing the computers was design for this then it should not be a problem. If not, it may be better to have the computers in groups and link them together via a single link. From davids description of the setup he wanted for his computers it appears he wanted a group of computers in his office and a group in the basement. It would therefore seem logical to me to link then using a subnet.

Just my 2 cent

Charlie.

Zyzzyx · Post by **Zyzzyx** » Sun Dec 14, 2003 3:14 pm

The reason I can see to have them on a different subnet is just classification neatness. Or if you have all of them going through a gateway separate from your main router (like the server for this farm, having two NICs).

You can have computers on the same subnet across different switches/hubs. The switch is just relaying the information of what IP addresses are behind it, and which IPs are located at each port (possibly more than one IP, if you have chained another switch off the first one).

haysdb · Post by **haysdb** » Sun Dec 14, 2003 4:19 pm

This is just a hobby for me also, so I could be completely full of it, but here is what I think...

Zyzzyx, what you say is true about being able to connect computers to any hub or switch on a network, independent of the subnet they are on, but...

I will have two DHCP servers, one serving the existing network and one serving the folding clients. For this reason I will have two independent networks, with the folding server operating as a bridge. The XP machines will pick up their IP addresses from a Dlink wireless broadband router. The folding clients will get their IP addresses (along with their OS) from the farm server.

If for any reason I can't get that to work, I will disable the DHCP server in the Dlink router and let the folding server assign IP addresses to all machines. The difference in wiring will be insignificant.

Physically the server will be upstairs and the farm clients downstairs. I already have two CAT5 runs between the office and a patch panel in an equipment closet in the basement next to a switch, with patch cables between the two, so configuring the network will just be a matter of plugging cables into different ports in the closet.

It sounds so easy now,

but I know how "delicate" the network configuration can be. I spent all night trying to get mine working again after simply relocating the folding@home folder on the server!

David

haysdb · Post by **haysdb** » Tue Dec 16, 2003 2:46 am

I did something really dumb. I ordered all the parts for a diskless client, except for a heatsink, since I have an Alpha heatsink on hand. Trouble is, the new motherboard doesn't have the holes around the CPU needed to mount the Alpha. So here I sit, ready to go, but short a heatsink.

I have ordered a batch of Arctic Cooling heatsinks from Directron for $9.99 each, but they won't be here until the end of the week. I might have to break down and buy one from compusa.

I'm becoming an "expert" on some of this Linux system administration stuff. Spend a couple of all-nighters trying every permutation of every solution offered anywhere within Google's reach, and you learn a lot, even though you get nothing at all to work!

I have Samba running. I am monitoring FaH on the folding server with Electron Microscope on my main Windows XP machine. I can print to the printer connected to the XP machine.
I can telnet into the Linux box from Windows. Now, if I could just find an X client for my XP machine, I could move the server to the basement. I assume there is such a thing, but haven't looked.
Folding@Home is running as a service on the folding server.
I am using cron to periodically set the file protection on the FAHlog.txt file that EMIII needs in order to monitor progress.

Someone somewhere talked about learning Linux via "total immersion" and that's pretty much what I have done, and am still doing. After all, I don't have any diskless clients running yet. At least now I feel I have the "infrastructure" necessary to get it to work, assuming it doesn't "just work" and I have to figure out why.

David

CharlieChan · Post by **CharlieChan** » Tue Dec 16, 2003 4:07 am

haysdb wrote: [*] I can telnet into the Linux box from Windows. Now, if I could just find an X client for my XP machine, I could move the server to the basement. I assume there is such a thing, but haven't looked.

www.cygwin.com

Charlie.