[Beowulf] Diskless Cluster Ram Disk Login
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Julien Leduc julien.leduc at lri.frMon Apr 23 13:05:16 PDT 2007
- Previous message: [Beowulf] Diskless Cluster Ram Disk Login
- Next message: [Beowulf] Diskless Cluster Ram Disk Login
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hello Stephen, it can be very frustrating to try to make a complete autonomous linux system directly on an autonomous box. So there are some clues: 1. try to build it in a controlled environment: virtual machine for example, or even a small chroot in your main environment to find out if you have all the devices, required libraries... 2. you can begin your work from a booting initrd + kernel: modify the initrd so that you stay in it, use tmpfs (or additional ramdisks), add network drivers, then you can format a bigger ramdisk (or use full tmpfs at the desired size mounted rw) and perform a pivot_root on it... You should find plenty of documentation looking for "embedded linux HOWTO", it will be easier for you since you should avoid all the cross compilation tricky part. For your ssh server you can try to use dropbear (smaller and specially designed for embedded systems), be sure you know what is done with dns: if your embedded system can resolve names? perform reverse on IPs? SSH can be picky when it comes to DNS questions. For my usage, I have a hacked rsh server modified to avoid any DNS checks (every communication is based on IPs), and executes any command as root on the embedded system. On big clusters RSH can cause some problems if you intend to keep more than 256 simultaneous connections from a single server. Regards, Julien Leduc snj78 at comcast.net wrote: > Hello everyone, > > I have built a diskless cluster using FC6 x86_64 and can get a node to > successfully boot up via dhcp and pxelinux. I am not using a NFS > because I want everything to reside on the node as to reduce network > traffic. Right now it just boots up to a text login screen and it won't > accept any login or remote logins at all. I have modified the rc > scripts in the root file system many times, even copying the whole /etc > directory straight over from the main PC, but it still won't allow any > logins. I was able to to edit the inittab file to 'wait' at the shell > /bin/bash, but I still couldn't communicate with the node. I am going > to try to install SSH, make sure the daemon sshd is running on the node > with pregenerated keys to allow logins without passwords, and see if > that will help any. All I want to do is be able to boot the node up and > run an HPL benchmark on it. What else do I have to do to the RAMDISK > root file system to communicate wit! h the h ead node and allow logins > via ssh, telnet, or even rsh so that I can get a benchmark? > Thanks in advance for any help. > > Stephen Jenkins > WSU > Beowulf Senior Project Group > > > ------------------------------------------------------------------------ > > _______________________________________________ > Beowulf mailing list, Beowulf at beowulf.org > To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
- Previous message: [Beowulf] Diskless Cluster Ram Disk Login
- Next message: [Beowulf] Diskless Cluster Ram Disk Login
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
