Hi, although my problem may be slightly out of topic, in my opinion the discussion about implementing/running a programm in a wider sense belongs somehow the the program itself. If you don't agree please point me to a more appropriate forum. I work in an agricultural research institute and (together with some colleagues) have to maintain a network with about 15 Unix machine (some SUNs, but mostly Linux) and a bunch of about 40 PC running DOS, Windows3.11, and Win95. * Some years ago, SUN propagated diskless clients booting over the net. But they soon stopped this and further on propagated at least 'dataless' clients with swap and a base OS on a local disk. So today, most SUNs returned to local machines. * 6 years ago we started to build up a network of DOS-PCs for office apps. We took the existing Unix network as the backbone and connected the PCs diskless (ok, they boot from a floppy, but have no HD) to a Unix machine with a program called PCSHARE. PCSHARE provides a TCP/IP stack for DOS so you can load Windows3.x from the initially mounted filesystems. This (more or less) worked good for pure MS-DOS and Windows3.x. With Win95 this does not work any more as we cannot boot Win95 diskless. So an increasing number of machines turn to local installations with all its drawbacks. * Linux machines more and more supersede the SUNs. Installing, running and maintaining two or three Linux machines is fun but when you are responsible for 10 or more it becomes a pain to keep track of all changes, updates, etc. This setup made me follow the discussions about booting over the network and configuring a whole set of machines from one server. The pure process of booting (broadcasting on the net, finding a boot server, getting the boot image, etc.) is one (solved) step, configuring and maintaining this system another. I am interested how you solve these problems with a (semi-)large LAN setup. Who can tell me about his experiences and dis-/advantages of the different ways of network-booting, especially concerning Linux: * X-Terminal provides a set of basic files, libs, and programs for each diskless client. You need about 20 - 30 MB diskspace for each. How do you manage to keep all these clients in sync with your server setup? * There is a solution (?) called transname from Thomas Schoebel-Theuer <Schoebel@informatik.uni-stuttgart.de>. Here is a short part of the announcement: > linux-2.0.21-transname.patch enables diskless clients, X-terminals etc to > mount the *root filesystem* of the server. This makes administration of > large pools *a lot* easier. > Currently different diskless clients must have their root "/" on different > directories on the server, beause each client has _some_ different > configuration files. However, most administration files (typically about 99%) > have the same contents on the clients and on the server, but have to be > (and maintained separately) just because of the 1% differences. Thomas solved this by implementing a mechanism of hostname translation into the kernel. You e.g. have files /etc/fstab#server1# /etc/fstab#client1# /etc/fstab#clientn# and directories like /var#server1# /var#client1# etc. These names will be translated by the kernel for the appropriate machine to the original names. In my opinion - a fascinating model. I started to install it but due to some problems and lack of time I couldn't finish it. The main disadvantage I see with this way is that it needs a kernel patch and I don't know if it sometimes will be integrated into standard kernels. Does anybody work with transname? * Today I found a URL for a package called Host Factory (http://www.wv.com). They write: > Host Factory lets you customize one prototype Unix system and use it to > build hundreds of similar systems. It seems to be a rcopy/rdist-based product. Is is a commercial product and I don't know if free equivalences exist. This is my overview about the different ways. Any experiences, hints and comments are very welcome. Thanks in advance Helmut -- -------------------------------------------------------------------- Helmut Lichtenberg (PGP Public Key available) heli@tzv.fal.de Institut f. Tierzucht und Tierverhalten - 31535 Neustadt - Germany --------------------------------------------------------------------
For requests or suggestions regarding this mailing list archive please write to netboot@gkminix.han.de.