netboot of tcl

[NewUser]

I don't know the Livebox, but most modems/routers have a builtin web server used to configure the device.  Find the dhcp server or default gateway on a working PC and plug that IP address (192.168.0.1?) in the address bar of your favorite browser.

[Juanito]

for the livebox, the default is 192.168.1.1 with username/password admin/admin

[cast-fish]

Hello

THat is a great help Juanto and Newuser.

I will follow that advice. 

The router here is about 8 years old. It was handed out new when i opened a DSL contract with Orange. I have been with that same company ever since on DSL until a recent discovery changed things for me.

Essentially i have several routers here from different company's.

Some routers are from British Telecom,
some routers are from Orange(now called "EE")
ALso i got a new router off ebay. It was clearly brand new and boxed but never ever worked. Inside it there was a manually inserted micro chip which was missing. I can only assume that these routers were factory errors. For a new router like that, It was extremely cheap to buy.....so

I hope to be able to get this tinycore "netboot" idea working using  a sister tinycore machine.
It is just useful to know and helps when dealing with old computers and such like (or thin clients)

Thx

V

[andyj]

PXE booting is based on DHCP and TFTP. If your home router is your home DHCP server then it won't provide the name and address of the TFTP server that will have the PXE image that the system you want to boot over the net so nothing will happen. You must turn off DHCP on your home router and set up your own DHCP server that will you can configure with the name and address of the TFTP server that has the PXE image. Usually these two are on the same system, but they don't have to be. Then put the pxelinux.0 file in the root of the TFTP server directory and make sure a pxelinux.cfg directory is in there. There you'll need a default file that looks something like this (change the directories and file names for your configuration):

DEFAULT pxe
LABEL pxe
KERNEL pxelinux.cfg/default-boot/vmlinuz
INITRD pxelinux.cfg/default-boot/core.gz,pxelinux.cfg/default-boot/mydata.gz
APPEND quiet noswap tce=/ httplist=192.168.1.1:/pxeboot.lst

As you can see you will also need an HTTP server (I use lighttpd) to serve up the list of tcz extensions and the extensions themselves.

I'm using this method now to boot multiple VM's in an ESX environment and I've used it in the past in the real world too.

[gerald_clark]

I suggest that instead of loading all your extensions using  http ( it was not designed for that ), you use an NFS share for your tce directory.

[andyj]

But then I'd need a NFS server and presumably a persistent connection. I don't have a problem with the NFS server per se, but I don't want a persistent connection. Once the VM boots I want it to be self sufficient and not depend on another server full time.

[gerald_clark]

Either it requires a network connection or it doesn't.
If you are going to PXE boot the workstation, it requires a network connection.
You already have a DHCP server, A DNS server, and a TFTP server.  Setting up an FTP server is trivial.
Unless you plan to PXE boot the machine, pull the cat 5 line and walk out of the room with the machine, I don't  see the problem.

[andyj]

I'm not following how the FTP server relates to the NFS server. But I'm not seeing the benefit of having the tcz's mounted into the local filesystem via NFS and thus making the NFS server a required always on part of every system attached to it versus letting the system download the tcz's it needs and then going on it's own.

[gerald_clark]

Then use the copy2fs flag.
The client machines do not need NFS server onboard.  Core has the nfs client built-in and does not need any client loaded to load extensions from an NFS server.

[curaga]

If you prefer a HTTP wget loop, you can do so in bootsync.sh, but gerald_clark is correct that httplist is too early to load many extensions properly.

[andyj]

I'm still not following why I should be having a problem. All the extensions get downloaded during tc-config. It's all working. What's broken about it?

[curaga]

Extension startup scripts may do things which expect the system to be up. Users to be fully setup, etc.

[gerald_clark]

Httplist was designed to load drivers to support a network based tce directory using NFS, NBD, AOE etc.  There is no dep handling, and high level services have not yet been started.

[andyj]

Fair enough. But wouldn't having the extensions on an NFS mount mean that every time something in an extension was accessed the system would have to go to the network to read the whole file just to get to the one file in the squashfs archive? Sounds inefficient.

[gerald_clark]

As I stated before, you can set the copy2fs flag, which will cause the extension to be copied to RAM when tce-loaded.