Hello,
I having issues and I'm not sure what they are so so finding a solution is proving very difficult. I had a cluster that worked fine when it was fedora 6 and debian 4 the boot server failed and was replaced with a fedora 7 system now it doesn't work.

The problems are as follows. There is a master node running fedora 7, this master is running a dhcp server which serves the other nodes with ip address. This part it does. It is supposed to boot the other disk less nodes over the network using pxe/tftp, this part it almost does. The problem is that the root for the nodes is mounted during boot via nfs but boot server will not let the nodes mount the nfs share. So the nodes boot so far an then stop saying NFS over TCP not available.

If I try to mount the nfs share just with another computer that has received its ip from the dhcp server but booted from its own hd, the mount times out unless I put the ip address and host name in the servers hosts file, then it works but this messes with the dhcp server.

Can someone help me sort this out I'm lost.

What I want is...
The boot server running fedora 7 to serve ip address to the other nodes and allow the to boot via pxe. I would like the nodes to connect to the internet through it using the servers other ethernet card that is connected to the rest of the network and has a ip address assigned by its mac address.

Thanks,
Phil

Okay, so I think that I have the dhcp server is almost working correctly, it gives out the correct ip to the nodes based on the hosts file and there mac address. The nodes now get the right initrd.img and kernel from tftpboot based on there mac address in the /tftpboot/pxelinux.cfg/ directory. However I am still getting an error from the nodes during the boot process...

nfs over tcp not available

and then the nodes just loops round trying to mount the nfs share.

I'm still not sure what the options for the following enteries in the dhcp.conf should be, or how to let the nodes get internet access through the boot server..

option domain-name-servers xxx.xxx.xxx.xxx;
option routers xxx.xxx.xxx.xxx;

the nodes also complain that eth0 doesn't have a default route set.

Needless to say that the nodes don't boot but I think I have made some progress.

The correct directory containing the root for the nodes are being exported...
/mnt/mini-itx-root 192.168.0.201(rw,async,insecure_locks,no_root_squash)

That is the correct ip for the node and it gets it from the server.

Can anyone help, I'm at the limit of my knowledge. I'm just a scientist this cluster is normally modelling plants.

Thanks,
Phil

Edit: rpcinfo -p shows that nfs is working over tcp...

program vers proto port service
100000 4 tcp 111 portmapper
100000 3 tcp 111 portmapper
100000 2 tcp 111 portmapper
100000 4 udp 111 portmapper
100000 3 udp 111 portmapper
100000 2 udp 111 portmapper
100000 4 0 111 portmapper
100000 3 0 111 portmapper
100000 2 0 111 portmapper
100024 1 udp 978 status
100024 1 tcp 981 status
100011 1 udp 694 rquotad
100011 2 udp 694 rquotad
100011 1 tcp 697 rquotad
100011 2 tcp 697 rquotad
100003 2 udp 2049 nfs
100003 3 udp 2049 nfs
100003 4 udp 2049 nfs
100021 1 udp 32823 nlockmgr
100021 3 udp 32823 nlockmgr
100021 4 udp 32823 nlockmgr
100003 2 tcp 2049 nfs
100003 3 tcp 2049 nfs
100003 4 tcp 2049 nfs
100021 1 tcp 54631 nlockmgr
100021 3 tcp 54631 nlockmgr
100021 4 tcp 54631 nlockmgr
100005 1 udp 710 mountd
100005 1 tcp 713 mountd
100005 2 udp 710 mountd
100005 2 tcp 713 mountd
100005 3 udp 710 mountd
100005 3 tcp 713 mountd

Okay I gave in because I couldn't find anything wrong, copied all the config files for dhcp, the exports and kernel and initrd.img etc onto a usb stick. Reinstalled Fedora 6. Replaced the config stuff, installed the right programs. Rebooted and it all works, apart from I haven't got the route to the internet from the nodes
setup. So if anyone can help me with that I would be thankful.

I have no idea why it didn't with Fedora 7 but I not going to worry about it at all.

Cheers,
Phil

Finished..
So for some reason the option in dhcp.conf for the DNS server(which I set as the ip of the router/dns server) was being ignored and there as an incorrect setting in resolv.conf which is why addresses could not be resolved. I'm not sure why the setting in dhcp.conf was not being used because when the address is served to the client it says that DNS1 is set and the value in there is right.

So it is all sorted out now... I know no-one has replied but the info might prove useful to someone with similar problems. I now have a full functional disless cluster for running my simulations and bootserver

Phil