> Here's another way to hang a system: > > 1. mount some nfs shares. > 2. shut down one of the machines serving the nfs share without first > unmounting it, and have a filemanager open to the remote filesystem > 3. Now try to shut down the client machine. It will hang on unmounting > the remote filesystem. (and now you learn to use alt-sysrq-r alt-sysrq-s > and alt-sysrq-b)
That's a good demonstration of one of the weaknesses of NFS. Unfortunately, for some of us, NFS is a necessary way of life. Imagine a system with a large central server (disk farm basically), feeding a network with 2,500 workstations, and over 5,000 registered servers. Each user must have access to any workstation, and have their home directory, and some data directories available. The only way to do that is that each of the clients is tied via NIS/NFS to the server. NIS serving the user names, and NFS mounting the home directories, and the data directories. NFS/NIS is a necessary evil. I'm not writing this looking for fixes. I've been running systems like that for years on AIX, and I'm just used to the headaches involved. I just brought it into the conversation because of the way civileme was "dismissing" NFS as to unstable to use. Boeing runs on NFS. As does Lockheed Martin, Northrop Grumman, AT&T Wireless, etc... You just have to approach it with the attitude that "It can, and at times will, wreak havoc". And it never disappoints. ;) JMHO-YMMV Ric
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com