knowledge on the level of the sorceres apprentice--newbee
have done lots of reading and homework on usiing clusterknoppic--found lots of confusing information
cthuling worse
not found any howtos to install
read tons of stuff on openmosix site
keep in mind that my goal is to see how simple it is to do

i would like to start a thread on getting it up and running

1. have 1 machine to be server dell poweredge 2300 518 megs memory tons of hd space dual pentium 2 400ghrtz
server has cable ports ( i assume are for networking--very new to this, manual somewhat confusing)
i have put a cat 5 cable into one of them and then into a hub--power light is on on the hub --no leds on poweredge.
placed clusterknoppix in cd and booted up
as expected no problems whatsoever.

2. second machine pentium 3 2.4 meghertz, 518 megs memory tons of disk space.
has a nic port on mother board--i have set the bios so it can boot from this port -- manual says it can boot from ethernet port
cat 5 cable from this port to hub

3. now what do i do?--server is running clusterknoppix off of cd other machine is not on

i really appreciate any help on this

The easy answer is burn another cd and boot it. Avoid the pxe/boot hassle.

Here are some general hints to get you to a pxe boot

On my computer, there are at least 4 places that need to be enabled in the bios to enable a network boot.
LAN,CDROM,HDRIVE is the boot order I use.

If you have a pci network card there is a small 3 pin cable that needs to connect to the motherboard in order to Wake on LAN and or NETWORK BOOT.
Also these should be enabled in the bios.

Run "/etc/init.d/openmosix restart" from a shell as root. Check for openmosix errors.
You will need to create /mfs and edit /etc/mtab and /etc/fstab for a hardrive install. Copy the proper mfs entry's from a live running cd.

Run /etc/init.d/terminalopenmosix restart from a terminal you will get a lot more info about what is happening.
You can mostly accept the given defaults.
Make sure you choose the right network module for the nodes.
pass any necesaary cheat codes to the nodes. My nodes need" xmodule=vesa" even running without a gui.
If all is well boot your node and you should see a knoppix start-up screen on your node.
johnb

After johnb minihowto (http://www.knoppix.net/forum/viewtopic.php?t=2461&postdays=0&postorder=asc&start=50),

edit the script /etc/init.d/knoppix-terminalopenmosixserver:

edit the line 333:
cp -a /dev/{null,zero,console,cloop,hd*,sd*,init*,kmem,fd*,loop*,mem,ram*,*random,pts,psaux,ptmx,std*,systty,sndstat,tty,tty[0-9]*} "${MINIROOT}/dev/"
to
cp -a /dev/{null,zero,console,cloop*,hd*,sd*,init*,kmem,fd*,loop*,mem,ram*,*random,pts,psaux,ptmx,std*,systty,sndstat,tty,tty[0-9]*} "${MINIROOT}/dev/"

(yes: only a '*' )

The problem: terminalopenmosixserver dont copy cloop0 device into miniroot. After node restart, cloop try link the nfs export dir to local /dev/cloop0, but this point dont exist.
And mount dont find the root filesystem...

For one, you should start the cluster-terminalServer so your 2nd computer finds something on the network to boot. You'll find that utility off of the Knoppix menu item under Services( IIRC ). Once you start the cluster-terminalServer, your server will now be ready to send out the ClusterKnoppix image to any machine on your network which is looking for a PXE boot server.

FWIW, here's a link to and IBM article on using and older version of ClusterKnoppix on systems with CDROMS. Informative for the newbie:

http://www-106.ibm.com/developerworks/library/l-clustknop.html

Is there no way to get a terminal server running with a knoppix-installed system? It seems all the howtos on here use a liveCD or a poor man's KNOPPIX/KNOPPIX image only for PXE to work....

BTW, is there to be a clusterKNOPPIX 3.7????

Regards,
AJG

The terminal server works fine as a hard drive install now. Thanks Joaquin for that "*".
You need 3 partitions for it to work:
1. hda1 normal knoppix-installer routine
2. hda2 This is the NFS mount for nodes. I do this as live-cd running the terminal server script. Make sure hda2 is mounted/writable.
3.Swap File

I assume there will be a CK-3.7. Seems to me computers will only be more inter-connected as processors become hotter/expensive.
johnb

Last year, around this time, I had my students investigate how to program a cluster so we installed QUANTIAN to the hd and set-up mfs on our LAN (25 Dell Optiplexs with PIV and 750MB RAM on fastETHERNET). We got demos like KANDEL and POVRAY to work, but we didn't have much success coding our own apps as we were trying to migrate JVMs (we had been studying OOP in java).

This year I'm considering hd installing ClusterKNOPPIX instead with mfs. Then we could use OOP in C++ and the fork() method to take advantage of the openMOSIX kernel. Anyone have success writing their own code like this?

I suppose there's always PVM or MPI and C....

Any idea if there will be a ClusterKNOPPIX 3.7 or 3.8?

Thanx,
AJG

You know what I'd really like to see?

Somw way of having a bunch of Knoppix boxes all running the open-source Firebird database server and working together to return result sets for other clients.

I have an application, talking to a Firebird server on my Windows box, that, because of my database design (an experimental design where "Customers" for example do not have their own customers table, rather a generic "attr_values" table, where each attribute (first name, last name etc) are stored as separate rows in this attr_values table........ yes, you might think that's really bodgy, but wow - it works great.)) tends to slow down a bit.

Anyway, the more rows I put into it, the *slightly* slower it takes to return records (using stored procedures to return a soert of cross-tab result set) and I was thinking that it'd be really good if several Firebird instances could get together and return records......... instead of just a single instance.

Firebird has a few good replicators, so this could really work. The Firebird people are looking at "clusters" for the version 3 release of Firebird (called Vulcan) but I thought I'd post this just to stir up some comments!

My thinking, in the interim, would be to have one Knoppix node acting as a co-ordinator - it receives connections from clients and passes those queries onto the first available Firebird node (running on another Knoppix box)..... it would get the results back and pass them back to the client. So basically, client A could ask for a recordset, and be served from database server node B. Client B could ask for another recordset, and be served from database server node C, instead of a single Firebird instance having to respond to and serve multiple clients.

Does this sound feasible?

Thanks - and my apologies if this is in the wrong forum....

Regards

Brad
Knoppix Newbie
http://phoenixsoftware.dyndns.biz/

hi Jorge, do u mean to use this for the DVD booting master node, or for the HDD installed master node?

thx